Private Equity and the firm-level impact: the case
for investments
Jorge Manuel Ribeiro Gomes
[email protected]
Master in Finance Dissertation
Supervisor:
Miguel Sousa, PhD
September, 2015
Brief Biographical Note
Jorge Gomes got its BSc. in Economics (at the time a 5-year long degree) from the
School of Economics and Management, University of Porto (FEP), back in 1997.
Since then he has worked in Corporate Banking, in Portuguese and International Banks,
having performed different roles, encompassing Coverage, Credit Risk and Product
positions.
Currently, he works as a Debt Finance originator for Global Corporates in an
International Team from a UK based International Bank, originating, structuring and
executing a full length of large (> €10m of ticket size) structured debt deals, spanning
bilateral and syndicated structures, pure refinancing and event-driven (M&A)
operations.
Mainly focused on the Iberian market, being the sole member of the team in Portugal,
he frequently works in deals across the whole International scope, from Italy and other
European countries, to UAE and India. He also works in close relationship with the
Investment Bank arm, namely the Debt and Equity Capital Markets teams, in order to
present the full spectrum of financing options to the Bank’s clients.
ii
Acknowledgements
To my wife, Ana, for enduring the long absences patiently and without whose support
this journey wouldn’t be possible.
My gratitude also goes to my brother, Miguel, for the invaluable crash course on
advanced VBA.
Last, but not least, to Prof. Miguel Sousa, for the support, from the very early beginning
of this research idea, with the useful suggestions, extreme availability and quickness in
answering to all my queries and doubts.
iii
Abstract
One of the most common anecdotal criticisms to private equity (“PE”) activity is that
they cut myopically capital expenditures. Notwithstanding the fact that it is relatively
clear that investments do fall after the buyout, it is far from answered the question on
whether this results from underinvestment or, alternatively, overinvestment - a
correction of an agency problem. Until recently, this has remained an understudied
subject in the literature, with an overinvestment correction hypothesis being implicitly
adopted, conditioned by the focus of research on US public-to-private deals and the lack
of private firms financial data in US keeping the debate on overinvestment of public
over private firms, opposing Jensen (1989) to Stein (1988), mainly on a theoretical
ground. Only more recently, this hypothesis started to be questioned with Sousa and
Jenkinson (2013), Bharath et al. (2014) and Ughetto (2014) concluding that evidence is,
at least, not supportive of an overinvestment explanation. At the same time, a recent US
study (Asker et al., 2015) also disputes the idea that public firms overinvest their private
counterparts. We analyse a sample of 92 PE entry deals that took place in Europe
between 2006 and 2010. We also compare a sample of 29 thousand European public
and private companies, during the last decade. We find some evidence, even though
limited, that PE impact negatively firms investment policies due to a mix of increased
financial constraints and probably to lower sensitivity to investment opportunities. In
any case, we found stronger evidence that the overinvestment correction is hardly a
valid explanation, as public firms, at least before the crisis, clearly invested less than
their private counterparts, under all matching criteria, and controlling for investment
opportunities and cash-flow. Under some specific matching criterion, they still did after
the crisis.
Key-words: private equity, investments, capital expenditures.
JEL-Codes: G11, G24, G34
iv
Resumo
Uma crítica frequente ao capital de risco (“CR”) é a de cortarem o investimento de
forma cega. Embora seja relativamente consensual que o investimento cai após a
aquisição, encontra-se por responder se isso resulta de subinvestimento ou antes de
sobreinvestimento (custos de agência). Até recentemente, a questão permaneceu
negligenciada na Literatura, com a hipótese sobreinvestimento a ser implicitamente
aceite, em parte pelo facto de a investigação ser centrada em transações de aquisição de
empresas cotadas nos EUA e devido à falta de dados financeiros de empresas não
cotadas nesse país, determinando que o debate sobre se as empresas cotadas investem
mais ou menos, opondo Jensen (1989) a Stein (1988), se mantivesse teórico. Apenas
recentemente, a hipótese de sobreinvestimento começou a ser questionada, com Sousa
and Jenkinson (2013), Bharath et al. (2014) e Ughetto (2014) a concluírem que a
evidência, não é favorável àquela explicação. Asker et al. (2015) apresentou também
nova evidência que contesta a ideia de que as empresas cotadas investem mais que as
não cotadas nos EUA. Analisámos uma amostra de 92 transações de aquisição de
empresas por CR, entre 2006 e 2010, e uma amostra de 29 mil empresas cotadas e não
cotadas Europeias, nos últimos 10 anos. Encontramos alguma evidência, embora
limitada, de que entrada de CR impacta negativamente a política de investimento das
empresas, devido a uma combinação de aumento de restrições financeiras e menor
sensibilidade às oportunidades de investimento. Em todo o caso, encontramos evidência
significativa de que a hipótese de correção de sobreinvestimento dificilmente será uma
explicação válida, tendo em conta que as empresas cotadas, pelo menos antes da crise,
investiam menos que os pares não cotados, em todos os critérios de formação de pares e
controlando as oportunidades de investimento e a rentabilidade. Sob determinado
critério, essa situação continuou mesmo depois da crise.
Palavras-chave: capital de risco, investimento, despesas de capital.
Códigos-JEL: G11, G24, G34
v
Index
Brief Biographical Note.............................................................................................. ii
Acknowledgements .................................................................................................... iii
Abstract ..................................................................................................................... iv
Resumo ........................................................................................................................v
Index of Tables and Figures .................................................................................... viii
1. Introduction .............................................................................................................1
2. Literature Review ....................................................................................................5
2.1. The role of Private Equity – early theoretical discussion ......................................5
2.2. Empirical Analysis to PE firm-level operating impact .........................................5
2.3. The Case for Investments ....................................................................................7
2.3.1. R&D Expenditures .......................................................................................7
2.3.2. Capital Expenditures .....................................................................................9
2.3.2.1. Overinvestment versus Underinvestment ....................................................9
2.3.2.2. Investment Cash Flow Sensitivities .......................................................... 11
2.3.2.3. Endogeneity ............................................................................................. 15
2.4. Discussion and Opened Questions ..................................................................... 16
3. Sample and Methodology ...................................................................................... 19
3.1 Sample ............................................................................................................... 19
3.1.1 Retrieving process and source ......................................................................19
3.1.2 Sample Description...................................................................................... 20
3.2 Methodological Considerations ..........................................................................23
3.2.1 Descriptive evolution ................................................................................... 23
3.2.2 Sector Medians ............................................................................................ 25
3.2.3 Investment Opportunities ............................................................................. 26
3.2.4 Investment Regressions ............................................................................... 27
3.2.5 Matching ..................................................................................................... 28
3.2.6 Endogeneity................................................................................................. 30
4. PE impact on Investments on Entry – Empirical findings ................................... 31
4.1 Descriptive Evolution......................................................................................... 31
4.2 Measuring Investment Opportunities .................................................................. 38
vi
4.3 Conditional Investment Intensities – PE firms vs. Peers ..................................... 40
5. The Public vs. Private discussion ..........................................................................46
5.1 The discussion and relevance for our analysis .................................................... 46
5.2 Empirical Findings ............................................................................................. 46
6. Conclusions ............................................................................................................ 52
Appendix I ................................................................................................................. 59
Appendix II ................................................................................................................ 60
Appendix III .............................................................................................................. 62
vii
Index of Tables and Figures
Tables
Table 1 - PE Entries sample “cleaning” ....................................................................... 20
Table 2 - Deal type breakdown .................................................................................... 22
Table 3 - Descriptive Statistics .................................................................................... 22
Table 4 - Predicting Tobin's Q ..................................................................................... 27
Table 5 - Investment Intensity change after entry......................................................... 33
Table 6 - Profitability and Cash-Flow change after entry ............................................. 36
Table 7 - Leverage and Interest Cover ......................................................................... 37
Table 8 - PE firms sensitivity to Investment Opportunities ..........................................39
Table 9 - Sensitivity to Investment Opportunities across PE and matched peers ...........41
Table 10 - Unconditional Investment Intensities .......................................................... 47
Table 11 - Conditional Investment Intensities .............................................................. 49
Table 12 - Main Studies addressing CAPEX impact on PE backed firms ..................... 59
Table 13 - Public and Private firms per Country/Legal Form ....................................... 62
Figures
Figure 1 - Target's Country of Origin and Sector ......................................................... 21
Figure 2 - Deals per year ............................................................................................. 21
Figure 3 - Kernel Density for CAPEX to TA (left) and Lag TA (right) on year n-1 ......30
Figure 4 – Median Cumulative Sales (left) and Total Assets (right) Growth ................ 31
viii
1. Introduction
In the late 80’s, as the first private equity (“PE”) wave neared its end, Jensen (1989), in
a seminal article, argued that leveraged buyouts (“LBO”) would emerge as a permanent
and superior form of organization, “eclipsing” public corporations.
Rappaport (1990) presented an opposite view, by considering LBOs a “cul-de sac”, due
to its self-limited nature, whose benefits, in terms of governance and agency costs
mitigation, could be matched by a permanent organization, the public corporation,
through other means.
The first significant empirical research on the subject was provided by Kaplan (1989),
who concluded that PE create value through LBOs, following significant improvements
in operating performance.
Other studies backed this overall conclusion. In one of the most recent and
comprehensive studies Guo et al. (2011) concluded that albeit that there are still some
improvements in operating performance of PE backed firms, they had, somehow,
reduced significantly during the second wave (late 90’s onwards).
In addition to the operating performance improvement, the studies also reported that
capital expenditures (“CAPEX”) typically fell after the buyout. This fact can be
consistent with two contradicting hypothesis (Wright et al., 2009): (1) post-buyout firms
are cash constrained and underinvest; and (2) the buyout governance structure induces
managers to reduce capital expenditures that are non value maximising.
While the latter would be a confirmation of Jensen’s free cash flow hypothesis, the
former could have some significant implications on the long term value of PE activity,
as an “artificially” lower CAPEX, to boost the buyout deleveraging and, thus, ensure to
the PE investor a higher return, could hamper the long term performance of the firm.
The question is, however, far from being straightforward. First, it is difficult to assess
whether a firm is postponing, or not, positive net present value (“NPV”) investments, as
the investment level data is hardly available. Second, the interpretation of the effect of
1
financial constraints on investment is not completely free from discussion in the
literature.
The more or less consensual evidence that CAPEX falls after the PE entry, was early
interpreted explicitly by Kaplan (1989), and implicitly since then, as a sign of a
correction of overinvestment due to free cash-flow / agency problems (Jensen, 1986).
This tacit assumption was somehow conditioned by the fact that a significant proportion
of the empirical research was focused on public-to-private deals and, as research is
prone to be US centric, the lack of private firms financial data has kept debate on the
overinvestment of public over private firms, opposing Jensen (1989) to Stein (1988) who claimed that public firms cut myopically investments due to short terms pressures mainly on a theoretical ground.
Hence, until recently, the impact of PE activities in firms’ investment policies has been
somehow a neglected topic, despite the fact that this is one of the most common
anecdotal criticisms to PE activity - the fact that they allegedly cut myopically CAPEX.
Only more recently, the overinvestment correction hypothesis started to be questioned,
as for the typical study (US/UK, large companies) evidence started to mount up that
there may be some flaws to the free cash flow hypothesis in explaining CAPEX
behaviour. Sousa and Jenkinson (2013), Bharath et al. (2014) and Ughetto (2014) have
concluded that there is some evidence that supports the underinvestment of PE firms or,
at least, that evidence is not supportive of an overinvestment explanation.
More recently, focusing on the public vs. private firm debate, Asker et al. (2015),
building on an exclusive and new US private firm database, showed that, in the US,
public firms invest less and are less sensitive to changes in investment opportunities
than private firms, even during the recent financial crisis, when private firms most
probably became more financially constrained than their public counterparts.
This dissertation seeks to explore this thematic, incorporating and combining some of
the early approaches (Kaplan, 1989) with the most recent contributions from the
literature, namely by adapting Asker et al. (2015) methodology to the PE context.
2
Using data collected from Zephyr and Amadeus databases, we analysed a sample of 92
European PE entry deals, that took place from 2006 to 2010, and c. 29 thousand
European public and private firms, during the last decade, with a twofold research goal.
First, we tried to assess the impact of PE entry in European companies during the last
decade, answering the question on whether PE backed firms cut myopically
investments, by comparing to its matched peers and controlling for variables commonly
used in the empirical investment literature to explain investment intensity.
Second, we also try to compare public to private firms investment policies, in order to
compare to Asker et al. (2015) results for US with Europe, and verify empirically the
question of public firm overinvestment (Jensen, 1986) vs. underinvestment (Stein, 1988)
hypothesis.
Both goals are related, as the common explanation for the investment intensity
reduction after the buyout is exactly the correction of overinvestment.
In a sample deeply marked by the financial crisis, which severely penalizes the
statistical significance of our results, we found limited evidence that investment of PE
backed firms falls below its peers, even after controlling for the variables that explain it.
This is only visible in a specific investment intensity metric and in relation to a certain
matching criterion, sector and return on assets (“ROA”), under which it is also possible
to conclude that, after the PE entry, firms become less sensitive to investment
opportunities.
For example, by year 3 after de PE buyout the PE firms’ investment intensity is lower
than its matched peers by sector and ROA in -2.2 pp and -10.6 pp, for CAPEX/Lag total
assets, CAPEX/sales, respectively. For the same matching criterion, the PE firm has
CAPEX/Lag total assets lower in 1.28 pp (sig. 10%) than its peers, holding investment
opportunities and profitability constant.
However, the fact that after PE entry firms become more financially constrained is quite
more robust to several specifications and samples. The critiques to investment cash flow
(“ICF”) sensitivities interpretation relates to the fact if they should, or not, be
considered a measure of the degree of financial constraints but do not apply to the
3
interpretation of a sign of its existence: financially constrained firms would have
positive and significant ICF sensitivities (Bertoni et al., 2013). In our sample only PE
firms show positive statistically significant ICF sensitivities in the post buyout period.
In any case, we found stronger evidence that European public firms invest less,
controlling investment opportunities and profitability, than private counterparts, under a
specific matching criterion (sector and ROA) or, at least they did before the crisis, under
all criteria. Although there are some differences, our results are broadly in line with in
Asker et al. (2015) findings.
Interestingly, the disappearance of the public firm underinvestment with the crisis,
under some matching criteria, seems consistent with a higher ICF sensitivity from
private firms, which as of common knowledge, have less access to capital markets and,
thus, probably became much more dependent on their internal financing sources, as
with the crisis the banking system financing availability shrunk.
This, of course, lacks further investigation, but seems also consistent with the fact that
the only period where unconstrained investment intensities become higher in public
firms than in (matched) private ones is during the start of the crisis period (2007/08),
and happens not because public firms increase their investment but because private
firms dramatically shrunk their CAPEX.
We also find possible that the different financing structure on the US and European
firms, with the former much less dependent on the banking system, can be a clue to
explain the differences from our results to ones found in Asker et al. (2015).
Besides this section, this report is structured as follows: in section 2 we review the
literature on the topic; in section 3 we discuss the methodological aspects of our study;
in section 4 we present our empirical findings regarding our sample of 92 European PE
large (>€50m) deals (entries) in the 2006/10 period; in section 5 we present a
comparison between public and private firms and; finally in section 6 we present our
summarized conclusions and suggestions for further research.
4
2. Literature Review
2.1. The role of Private Equity – early theoretical discussion
Jensen (1989), in a seminal article, argued that LBOs would emerge as a permanent and
superior form of organization, “eclipsing” pubic corporations.
Building on his earlier concept of the disciplining role of debt as a mitigator of agency
costs raised by free cash flow (Jensen, 1986), the author argued that together with the
leveraged (more efficient) capital structure, LBOs enjoyed a concentrated ownership,
resulting in closer monitoring of the managers and stronger managerial incentives.
These unique features, he argued, enabled LBO managers to add value more effectively,
otherwise wasted by public corporations, which could be glimpsed by the 50% average
premium paid (at the time) in LBOs.
On the opposite side, Rappaport (1990) argued LBOs were an economic “cul-de sac”,
due to the self-limited nature: by design, LBOs are a transitory organization, as the
limited-partnership agreements provide ten-year duration for the partnership. Thus, in
order to maximize returns, the sponsor needs to generate cash from operations and
divestitures to reduce debt toward pre-buyout levels, in order to cash-out, either by
returning the company to public markets by selling it to a strategic buyer.
At most, he argues, the LBO is a short term “shock therapy”, but whose benefits in
terms of governance and agency costs mitigation could be matched by a permanent
organization, the public corporation, by other means.
2.2. Empirical Analysis to PE firm-level operating impact
In the same year that Jensen (1989) made his proposition, Kaplan (1989) produced the
first comprehensive insight on the firm-level impact of PE activity1. By analysing 76
large management buyouts (“MBOs”), between 1980 and 1986, and comparing
operating income and cash flow variables evolution 1 year pre-buyout with up to 3
1
The author refers the study as the first to use pre buy-out as well as post buyout data.
5
years after buyout, controlling for divestures, differences in growth, and industry, he
found that buyout firms significantly outperform non-buyout firms 2.
The author analysed the causes of this outperformance and tested three hypotheses: (i)
employee-wealth-transfer hypothesis; (ii) information-advantage or underpricing
hypothesis; (iii) reduced agency cost or new incentive hypothesis. The author concludes
that evidence points to the fact that the operating improvements are generated by
incentives rather than wealth transfers from employees or superior managerial
information.
Since then, while mainly focused in US and UK, several studies analysed the operating
impact of PE in the buyout firms, and found consistently overall results. Strömberg
(2009) provides a useful summary of related studies and findings.
We believe it’s worth highlighting Guo et al. (2011), due to both its breadth and
comparability3 with Kaplan (1989). The authors analysed the operating performance
and value creation on the second PE wave, by studying a sample of 192 LBOs, spanning
from 1990 to 2006.
They found that, unlike what was documented in relation to the first wave, gains in
operating performance are comparable, or, at best, slightly higher than those observed
on industry (and matched pre-buyout characteristics), depending on the measure4.
Besides the studies on operating performance impact, several other studies have been
focusing on several specific issues5.
2
For example, operating income (EBITDA) in buyout firms increased a median 15.3%, 30.7% and 42%,
in years +1, +2 and +3 after the buyout, when comparing to pre-buyout year, significant at 1% level. The
difference to median industry change was -2.7%, +0.7% and 24.1%, also significant, meaning that the
buyout firm performance is more or less the same as the industry in the first two years after the buyout,
but in the third year clearly t outperforms its non buyout peers. Similar results were found in net cash
flow (EBITDA-CAPEX), and in these two variables as a percentage of sales and assets.
3
Both in terms of certain methodologies used as well as the direct comparison the authors make several
times along the text.
4
For example, the industry-adjusted changes - comparable to prior research – both for EBITDA to sales
and net cash flow to sales do not show any major gains. Using the industry performance and market-tobook-adjusted change, there they found a significant increase from year −1 to year +1 or +2 for both
EBITDA to sales and net cash flow to sales. Still, even in these cases, the magnitudes are substantially
smaller than reported by Kaplan (1989).
5
Gilligan and Wright (2012) provide useful Tables with main research articles on several PE related
subjects. Wood and Wright (2009) also provide an extensive compilation of the studies on the effect on
employment and wages.
6
2.3. The Case for Investments
One of the least studied aspects of PE impact regards investments. However, if we split
investments in to two strands, the research and development (“R&D”) expenditures,
usually referred to as a proxy to long-term investments, and capital expenditures, the
results and challenges posed by the existing research are quite different.
We’ll forgo R&D expenditures, on the back of, among other considerations, the
different accounting treatments under different GAAP, which can make an analysis
based on several countries complex.
Nonetheless, we believe it’s worth briefly
mentioning the main conclusions regarding R&D.
2.3.1. R&D Expenditures
R&D expenditures are more studied than capital expenditures. However there seems to
be fewer consensuses on the outcome.
Studying 131 LBOs between 1981 and 1986, Lichtenberg and Siegel (1990) found data
inconclusive regarding the impact on R&D. First, firms involved in LBOs were much
less R&D intensive (mean R&D to sales 1%) than other firms (3.5%) before the LBO.
Second, the difference is larger in the three years after than in the three years before the
buyout. However, the change in these differences was not statistically significant. Third,
the relative R&D intensity of LBO firms appears to have been declining in the years
before the buyout, despite not statistically significant.
Long and Ravenscraft (1993) confirm that LBOs target significantly below normal
R&D intensive firms. Pre-LBO R&D to sales is less than half the overall manufacturing
average. Nevertheless some LBOs did occur in high tech industry, as the large variance
in R&D intensity among LBO firms denoted.
Unlike Lichtenberg and Siegel (1990), however, they found a significant and dramatic
decline in R&D intensity as a result of the buyout: 40%. Still, they also find that the
declines in R&D intensity do not appear to hurt the ability of LBOs to generate
performance gains. On average, LBOs improve operating performance by 15 percent or
more. Cutbacks in R&D have no statistically significant effect on performance.
According to one of the estimated equations, for a typical R&D intensity cut of 0.63,
7
performance would decline by 0.19, which is only about 10% of the 2.06 increase in
cash flow/sales.
The theoretical foundations the authors present to indicate the reduction in R&D was
expectable after the LBO are more or less in line with Hall et al. (1990), who
established a clear relationship between leverage and reduced R&D spending. The
authors argue that R&D could be an unintended victim of the trend to shift financing
towards debt.
Nevertheless, they considered they were unable to demonstrate that the projects that
were eliminated in LBO restructurings were worthwhile (high social or private returns).
In a more recent work, Lerner et al. (2011) criticised R&D expenditures as a measure
for long term investment in innovation. According to the authors, not all research
expenditures are well spent, and some critics suggest that many corporate research
activities are wasteful and yield a low return, making changes in R&D expenditures
difficult to interpret. The authors used an alternative measure, patenting data, and found
no evidence that PE backed firms changed their patenting origination pattern.
Furthermore, they concluded that those firms become more cited.
Against this idea, Chauvin and Hirschey (1993) identify a clear and significant
relationship between R&D spending intensity and corporate value, although varying
with size, for several sectors.
Ughetto (2010) cites several other studies that support of the hypothesis that PE
intervention is not detrimental to long-term investments in R&D and innovation. This
author considered that the use of R&D expenditures or more direct measures of
innovative output such as productivity growth might prove more useful, than patents,
for a more complete assessment of the difficult task of measuring innovative effort.
Nevertheless, the key takeaway from the author’s findings is the suggestion that
evaluating all buyouts on the same performance metrics without taking into account the
characteristics of the investors/deal may not be appropriate, as they seem to have some
impact on the innovation performance of firms.
8
2.3.2. Capital Expenditures
The question with capital expenditures is not whether the CAPEX falls after the buyout,
or not, as there seems to be a relative consensus in this conclusion6.
Albeit only few studies refer to CAPEX directly, this assessment can be drawn
indirectly in most of the studies on the impact on operating performance, by comparing
the conclusions on EBITDA versus net cash flow evolution, two variables commonly
analysed. As it is usually observed net cash flow increasing more than EBITDA, and
taking into account the difference between the variables is CAPEX, the assertion seems
to be self explanatory.
A summary of the main studies addressing the PE/CAPEX topic is provided in
Appendix I.
2.3.2.1. Overinvestment versus Underinvestment
The real question is the fact that the CAPEX reduction can be consistent with two
contradicting hypothesis, a underinvestment hypothesis and a overinvesting hypothesis,
or, as Wright et al. (2009) describe it: (1) post-buyout firms are cash constrained and
underinvest; and (2) the buyout governance structure induces managers to reduce capital
expenditures that are non-value-maximising.
While the latter would be a confirmation of Jensen’s free cash flow hypothesis, the
former could have some significant implications on the long term value of PE activity,
as an “artificially” lower CAPEX, to boost the buyout deleveraging and, thus, ensure to
the PE investor a higher return, could hamper the long term performance of the firm.
The question is, however, far from being easy to answer. One of the reasons is that it is
difficult to assess whether a firm is postponing, or not, positive NPV investments, as the
investment-level data is hardly available.
Kaplan (1989), analysing the evolution of market adjusted return to shareholders,
concludes that as the MBO firms provide their shareholders with a higher return than
6
Exceptions made to Boucly et al. (2011) that found that in a sample of 839 French deals firms increase
CAPEX and Engel and Stiebale (2014) which found that in a sample of UK and French PE backed SMEs,
investment increased.
9
the market, there is indirect evidence that the reduced CAPEX was referring to value
decreasing projects and, thus, the reduction increased the firm value.
On the opposite direction, Sousa and Jenkinson (2013) show that PE backed firms
exited through IPO increase much more the CAPEX than firms that exited through
secondary buyouts (“SBO”). If the overinvestment hypothesis was right, and IPO firms
started to invest in value decreasing projects, they shouldn’t be able to significantly and
substantially outperform the market.
As IPO firms invest more than firms than remain under PE hands, and the abnormal
return suggests they invest in value increasing projects, this evidence, albeit indirect,
seems to imply that PE backed firms do postpone value increasing investments.
Interestingly, although not addressing the overinvestment/underinvestment discussion
directly, some years before Holthausen and Larcker (1996) had already casted some
doubts on the overinvestment correction, as a justification for the lower CAPEX in PE
backed firms.
They found that, for at least four years after the IPO, reverse LBO (“RLBO”) firms
outperform their industries on an accounting basis performance but experience a
performance decline. The authors also found that they spend significantly less than the
“industry norm” on CAPEX in the year prior to the IPO, but this difference disappears
afterwards.
Furthermore, they found no relationship between the changes in CAPEX and leverage
and managerial ownership. However, non-managerial insider ownership was found to
be significantly and negatively related with changes in CAPEX. The results are not
changed by including a variable to control for the IPO proceeds used to retire debt, as
many firms explicitly indicate that they are going public in order to raise funds for
increasing CAPEX.
It is argued that if the RLBOs are constrained in their ability to make investments, their
sample is unlikely to exhibit the positive incentive effects associated with debt as
described by Jensen (1989), as those apply to firms with free cash flow generating
ability and no profitable investment opportunities.
10
The authors considered the finding of an increased CAPEX after the RLBO as
consistent with the firms being cash constrained before, in the case of the absence of an
exogenous shock (for example a change in investment opportunities) to justify the
behaviour.
Chung (2011), analysing a UK buyout sample, suggests that the overinvestment
correction only happens when the target is a public firm and when it’s a private firm the
PE acts to increase value by reducing financial constraints.
More recently, Bharath et al. (2014) analysed a considerable sample of going private
(PE buyout, MBOs and operating firm) plant-level detail transactions, spanning from
1981 to 2005.
The authors found that, relative to control groups (industry-age-initial size groups),
companies decrease CAPEX after going private, which suggests that public firms invest
more than comparable private firms, and would traditionally be considered as a sign of
overinvestment due to agency problems leading to “empire building”.
However, they argue that if firms had been overinvesting, when they were public, they
would have expected to see an improvement in productivity afterwards, but they found
no such evidence, relative to control groups.
In fact, they argue that going-private firms achieve the productivity improvements not
by improving the productivity in individual plants but by selling the low productivity
ones. They conclude the data does not support the overinvestment thesis.
2.3.2.2. Investment Cash Flow Sensitivities
The logic behind an underinvestment hypothesis is the fact that PE backed firms may
face cash flow constraints, on the back of heavy debt repayments and associated
restrictions7.
7
In fact it is common for LBO structures to have considerable and direct restrictions on CAPEX and
acquisitions as well as maintenance covenants. See for instance the Loan Market Association
(http://www.lma.eu.com/default.aspx) or S&P Loan market guide (https://www.lcdcomps.com/).
11
In fact, Achleitner and Figge (2014), found that financial buyouts (SBO) use 28–30%
more leverage, measured in terms of Debt/EBITDA, than other buyouts, even after
controlling for debt market conditions at the time of the transaction 8.
However, the relationship between financial constraints, cash flow and investment is not
completely free from discussion in literature.
In a seminal study on the subject, Fazzari et al. (1988) show there is a link between
financial factors and investment decisions. The theoretical foundation for the research
was the notion that unlike in Modigliani-Miller world, factors such as transaction costs,
tax advantages, agency problems, costs of financial distress, and asymmetric
information, make internal finance less costly than new shares or debt.
The authors then argue that, under this reasoning, firms that pay higher dividends are
less sensitive to variations in cash flow than firms that exhaust nearly all of their
internal cash flows.
This conclusion was challenged by Kaplan and Zingales (1997) who found firms with
low dividends and that could invested if needed and were far from being cash
constrained. Fazzari et al. (2000) and Kaplan and Zingales (2000) continued the debate
and Moyen (2004), in a tentative reconciliation approach, showed that the classification
of a firm as under financial constraints is hard and results vary under different criteria.
Nevertheless, investment-cash flow (ICF) sensitivities, in the logic of Fazzari et al.
(1988), continue to be widely used by scholars as a theoretical framework.
ICF sensitivity can be defined as the impact that variations in a measure of cash flow
have in the investment intensity, and in the investment literature are usually the
coefficient of a regression of an investment intensity measure (e.g. CAPEX/Total
assets) against a measure of cash flow (commonly EBITDA/Total assets).
Bertoni et al. (2013) suggest that ICF sensitivity should not be used as a direct signal of
the severity of financial constraints but as an indicator of the existence of financial
constraints: the ICF sensitivity will not be significantly different from zero for non8
However, unlike Sousa and Jenkinson (2013) they found no robust evidence that financial buyouts have
lower equity returns than other buyouts or offer less potential for operational performance improvements.
These authors do not address the CAPEX issue.
12
financially constrained firms, but a positive and significant ICF sensitivity would
indicate the existence of financial constraints. According to the authors, the Kaplan and
Zingales (1997) critique refers to the monotonicity of the relationship not about its
sign.
Using an ICF sensitivity approach, Engel and Stiebale (2014) analysed a sample of UK
and French SME’s and concluded that PE enhance investment and reduce financial
constraints.
Interestingly, smaller deals, typically the ones that would capture SMEs, are usually
excluded from the sample in most studies we reviewed, on the back of less disclosure
and information availability. The authors refer to the fact that they believe SMEs are in
general more cash constrained, which today is under a generalized public debate in
Europe. This can determine a whole different logic behind the PE intervention and,
hence, make the results less comparable to the other studies.
The same reasoning can be applied to Chung (2011), as despite a broader sample, has a
median deal value of £10m, for the private-to-private subsample, signalling that he
captured a significant proportion of small companies, hence probably influencing his
overall conclusion that in private-to-private deals the PE intervention alleviates
financial constraints.
Ughetto (2014) showed that the characteristics of the deal, namely jurisdiction,
applicable law and of the lead PE investor can impact target firm’s investments and ICF
sensitivities.
Interestingly, this conclusion is highly coherent with the fact that the only studies, found
in our review, that conclude for a positive impact in investment in PE backed firms,
Boucly et al. (2011) and Engel and Stiebale (2014), relate to a particular market,
France, and to a usually out of scope type of firms, SMEs.
Regarding the French market, Ughetto (2014) also addresses its peculiarities in term of
Legal System. This relates with a strand of the literature usually referred as “Law and
Finance”, in which is shown the impact of the legal system in valuations – see for
instance La Porta et al. (2002).
13
An additional difficulty, relating the ICF sensitivity approach, is the fact that a positive
ICF sensitivity can be viewed both as a underinvestment or a overinvestment symptom,
depending on whether you accept Myers and Majluf (1984) asymmetrical information
hypothesis or Jensen (1986) agency / free cash flow theory.
In fact, if we believe that the asymmetrical information will lead to an inflated external
funds cost, the positive ICF sensitivity will lead the firm to pass on positive NPV
projects and, thus, underinvest. On the other hand, if we believe that in managers’ mind
internal funds are too inexpensive, the agency / free cash flow hypothesis, they will tend
to overinvest.
However, if ICF sensitivity typically leads to overinvestment, then as managerial
ownership increases, this sensitivity should decrease, as agency related issues started to
disappear. Nevertheless, an initial finding of Morck et al. (1988) showed that
managerial ownership has an “entrenchment effect” from a certain level and, thus, does
not vary monotonically.
Hadlock (1998) built a non linear model between management ownership and ICF
sensitivities which deals with the “entrenchment”9 effect. The author concludes that his
findings
are
consistent
with
asymmetric-information
problems
(hence,
the
underinvestment interpretation) becoming more severe as managers care more about
shareholder value, which is backed by the fact that the relationship between ownership
and ICF sensitivities is strongest for the highest Tobin’s Q firms, the commonly used
proxy for growth/investment opportunities 10.
Despite the debate on ICF sensitivities as a measure, or not, of financing constraints, the
investment intensity regressions can still be used while being agnostic on the financing
constraints interpretation (Asker et al., 2015).
Furthermore, regardless of our position in relation this question, it is of common
knowledge that LBO structures face restrictions on investments, acquisitions, either
9
Weisbach (1988) defines entrenchment as “Managerial entrenchment occurs when managers gain so
much power that they are able to use the firm to further their own interests rather than the interests of
shareholders”.
10
Usually Tobin’s Q is defined as the market value of the firm divided by the replacement value of its
capital stock
14
directly or through restrictive covenants11. Hence it is hardly debatable that they face
some degree of constraints.
2.3.2.3. Endogeneity
Another important issue, which requires attention when addressing the CAPEX issue, is
the endogeneity question. It is possible that PE could know beforehand those companies
that require less CAPEX or have lower positive NPV investment opportunities and that
fact determines by itself the posterior evolution.
The lower CAPEX can be a symptom of less investment opportunities. In fact, as
Muscarella and Vetsuypens (1990) conclude “high debt usage typically found in LBOs
may not be an appropriate financing structure for companies with large capital
expenditures needs.”.
Indeed, Bharath and Dittmar (2010) built 2 models, Cox hazard and a logit, to predict if
a firm will go private12 and the coefficient for the variable CAPEX / Sales is negative
both for MBOs and PE buyouts, albeit only statistically significant in MBOs.
The use of matched samples and firm fixed effects, which absorb all not visible
differences, mitigate this issue.
Some authors enhance the approach to endogeneity with an Instrumental Variable
approach (Asker et al., 2015) or a Treatment-Effect model (Sousa and Jenkinson, 2013),
either as a complement or robustness verification for the matching procedure, using an
exogenous control for a variable that affects the company status, without directly
affecting investment.
Others use a propensity score, as in Bharath et al. (2014), where following previous
work from two of the authors (Bharath and Dittmar, 2010) they include an additional
match criteria based on the probability (in the case) of going private. By matching on
the propensity score, they compare the results for the establishments that went private
with the ones exhibited by firms that had a similar probability of being selected into
going-private event.
11
12
Common examples are Debt/EBITDA or even Debt/(EBITDA-CAPEX) and EBIT/Interest.
They attain accuracy rates higher than 80%.
15
2.4. Discussion and Opened Questions
Despite the fact that in the last two years the theme started to receive some attention, we
can still consider there is a limited the scope of literature regarding to the impact of PE
in investments. This seems to contrast with the importance of the topic.
A report produced by Frontier Economics (2013) for the European Private Equity &
Venture Capital Association (EVCA) stated “We do not yet know enough about the
incremental impact of private equity on fixed capital formation (...)”
There are two main reasons for the importance of the subject. First, if it was verified
that PE backed firms cut necessary, positive NPV investments, this could, to some
extent, undermine the previous operating performance gains conclusions. In the words
of Smith (1990) “(...) cutbacks in capital expenditures, are alleged to compromise the
long-run competitive position of the firm in order to increase short-run cash flows.” 13.
To illustrate this point, imagine a company engaged in manufacturing one good, by
using just one machine, with very expensive parts and components. If the firm cuts the
maintenance CAPEX it could be able to attain a significant cash flow for some time,
higher than its peers, who keep replacing defective or worn out parts, or upgrading
some of its components with more technologically advanced solution. At some point,
the temporary gain on cash flow will start to dent the firm’s profitability and
productivity, as stoppage times and expensive repairs start to happen.
Given the length of years usually covered by empirical studies (up to 3 years postbuyout), this seems, at least theoretically, an admissible scenario.
This argument could be the explanation for Holthausen and Larcker (1996) lagged
performance reversal after the IPO/RLBO. Bruton et al. (2002), who confirm these
results14, refer they would expect a firm to converge to a typical industry firm after the
IPO, suffering again from the agency issue. However, they expected it to happen more
quickly and not to be such a lagged effect. They conclude that “efficacy of agency
13
Albeit he concludes that the cuts in CAPEX are not responsible for the short-term increase in operating
cash flows, as CAPEX is a non-operating use of cash, the point still remains open for the long run.
14
On the operating side; they don’t address the CAPEX issue.
16
theory for explaining a complex topic such as firm performance during the buyout cycle
may be limited”.
Second, this is, in fact, one of the most common anecdotal criticisms made to PE
activity: “The most common criticism of private equity activities claims that such funds
apply a short-term calculus (...) which in turn strongly implies that capital spending
should decline or at a minimum underperform other peer companies” (Shapiro and
Pham, 2009) .
To some extent one can interpret the, until recently, apparent lack of interest in the
literature for the theme as an implicit acceptance of the evidence that the PE firms cut
CAPEX after the buyout as a confirmation of the correction of the agency problem
(Jensen, 1986).
However, this tacit acceptance seems somehow disconnected from the broader debate
on whether public firms really overinvest on the back of agency /free cash flow
problems.
Stein (1988) presented a complete opposite view, by describing what he called as a
“managerial myopia” that lead managers in public corporations, to sacrifice long term
goals and investments, to be able to present steadily growing quarter earnings and thus
prevent takeovers. Hence, under this view, public firms should underinvest.
Given the fact that research is prone to be US centric and the absence of private firms’
data in the US15, the question remained mostly as a theoretical debate. Nevertheless,
some surveys seemed to back Stein (1988) assertion, by pointing to the fact that public
firm managers prefer short term horizon investments, believing that investors fail to
properly value long term investments (Poterba and Summers, 1995) and that managers
would avoid engaging positive NPV projects if that implied an impact in current
quarter’s earnings (Graham et al., 2005).
Recently, building on a exclusive and new private firm database, Asker et al. (2015)
show that, in US, public firms invest less and are less sensitive to changes in investment
15
Unlike in Europe, in US it is not mandatory for private firms to deposit/publish their financial
statements.
17
opportunities than private firms, even during the recent financial crisis, when private
firms most probably became more financially constrained than their public counterparts.
On a different approach, Kerstein and Kim (1995) had already shown that, after
controlling for parallel earnings information and size-related pre disclosure information
differences, CAPEX changes are strongly and positively associated with excess
returns16.
These findings support Stein (1988) idea of underinvestment in public firms due to
“myopia” rather than Jensen (1986) notion that public firm are prone to overinvestment
as a result of free cash flow / agency related problems.
This can bring some additional light to the investment impact debate on PE backed
firms. If in fact, the majority of studies show that PE impact negatively investment, and
if public firms tend to invest less than private counterparts, it seems reasonable to, at
least, suspect that this reduction in investment is not related to a agency problem
correction but to a strategic approach from the financial sponsor to release as much cash
as possible to meet debt repayments and de-lever17.
Albeit some studies have recently debated the investment issue directly, there seems to
be still significant space to research. Engel and Stiebale (2014), for example, focus on
SME’s, hardly were the main discussion is centred. Bharath et al. (2014) use plant level
data (US only), which can be at the same time information rich but also somehow lead
to the firm broader picture loss and its hardly replicated and comparable to previous
studies.
Finally, Ughetto (2014) narrows the analysis in private to private transactions, low and
medium tech firms and in 4 countries. Furthermore, this author uses an Euler equation
to deal with the investment opportunities issue (Q theory), instead of the more widely
used proxies such as sales growth or industry Q (Asker et al., 2015).
16
These results are in line with previous works such as McConnell and Muscarella (1985) who found
empirical evidence, from market reaction to unexpected decreases and increases in CAPEX, more
consistent with market value maximization rather than size maximization.
17
Please bear in mind that we are referring only to CAPEX and are not questioning the managerial
incentives and ownership control impacts on performance, which are a separate discussion to which we
are agnostic as far as this study goes.
18
3. Sample and Methodology
3.1 Sample
3.1.1 Retrieving process and source
Bearing in mind that our goal is to analyse the PE first time entry impact 1, our sample,
retrieved from Zephyr database, encompasses all PE entry deals (“Take Private” and
“Vendor Sale”), majority stakes, that were not secondary but-outs, in all sector with
exception of banking and insurance or holding companies/head offices, in the 2006 to
2010 period, with a minimum €50m (50 million Euros) deal value and in European
Union (“Euro-28”).
The end date is justified by the fact that we need 3 years of post deal financials for our
analysis and, at the time of retrieval, the last year with financials was 2013. The start
date is conditioned by the fact that Amadeus only provides 10 years of financials and we
need 2 years of pre-del data, in order to have beginning of year values for the lagged
variables in the pre-deal year.
As for the deals size, the choice is justified, as usual, in order to exclude smaller
companies and deals harder to analyse, or at least be less comparable with larger ones.
Bank and insurance are not comparable to other non-financial firms, and fall out of this
study scope, and regarding “Head Offices” or “Holding” activities it is commonly
accepted that these companies are difficult to analyse on a non case study basis.
Our query resulted in 585 deals, which we reduced to 443, mainly due to the lack of a
BvD ID number which precludes us from retrieving financials in Amadeus.
As shown in Table 1, for 164 companies there was no data (no retrieval or retrieved
fields with “n.a.”.
Finally, for the 279 companies for which we had data, we eliminated all the companies
(187) were we didn’t have all our key metrics, for all the years. This envisaged having a
1
Secondary or posterior Buy-Outs impact and motives can be very different from the first entry and
should be analysed separately. See, for example, Sousa and Jenkinson (2013) or Achleitner and Figge
(2014) for this topic.
19
sample highly comparable through the years (balanced panel). We ended up accepting
some lacunae in data for secondary metrics, namely debt, EBIT and interests.
Table 1 - PE Entries sample “cleaning”
Zephir Sample
Take out
585
Insolvency
Administration
Public takeover - Unsuccessful
Acquisition Increase
Unknown BvD ID
1
2
18
10
111
Sample for financials retrieving
443
No Data
Incomplete Financials
164
187
Final Sample
92
The size of our sample is far from being uncommon, as it can be seen in Appendix I.
Studies with large samples usually relax the deal size threshold or are specifically
targeting to analyse SME’s, which as we already addressed, can vary from larger
corporates both in the deal drivers as well as in posterior impact.
Furthermore, as we show further below, the elimination of incomplete financial firms
does not seem to have any statistically significant impact in our pre-deal year metrics.
3.1.2 Sample Description
Our 92 target companies are from 11 of Euro-28 countries, notwithstanding the fact
that, as one would expect, the majority (50%) is from the UK.
In terms of sectors, our sample encompasses 76 different 4-digit Nace codes, but to give
a broader view of the different activities we grouped them in EVCA’s sector clusters2,
which show that 63% of our sample comes from the Business and Industrial Products
and Services (“BIPS”) and Consumer Products, Services and Retail (“CPSR”), leaving a
residual importance for tech related companies, utilities and other sectors. Our sample
seems to be dominated by “old economy” or traditional sectors.
2
Please refer to EVCA website for details of correspondence between clusters and Nace codes:
http://www.evca.eu/media/12926/sectoral_classification.pdf.
20
Figure 1 depicts countries and sectors breakdown in detail.
Figure 1 - Target's Country3 of Origin and Sector4
Targets' Country
Sector - EVCA clusters
46
34
24
18
10
10
3
1
2
6
2
1
1
4
2
6
3
ACM BIPS CPSR
BE CZ DE ES FI FR GB HU IT NL SE
11
EE
FS
ICT
LS
Concerning deal years, as depicted in Figure 2, our sample is dominated by pre and
early start of the crisis deals as 64% of the transactions occurred in 2006/7.
Figure 2 - Deals per year
29
30
18
10
5
2006
2007
2008
2009
2010
Finally, Table 2 shows that the majority of the deals in our sample were institutional
buy-outs, and our sample includes both private to private and public to private deals.
3
Countries ISO Codes: Belgium (BE), Czech Republic (CZ), Germany (DE), Spain (ES), Finland (FI),
France (FR), United Kingdom (GB), Hungary (HU), Italy (IT), Netherlands (NL), Sweden (SE).
4
EVCA Clusters: Agriculture, Chemicals and Materials (ACM), Business and Industrial Products and
Services (BIPS), Consumer Products Services and Retail (CPSR), Energy and Environment (EE),
Financial Services (FS), Communications Computer and Consumer Electronics (ICT) and Life Sciences
(LS).
21
Table 2 - Deal type breakdown
Institutional buy-out 100%
Institutional buy-out 50%-99.9%
Acquisition 100%
Acquisition 75% minus one vote
Management buy-out 100%
57
20
5
1
9
As shown in Table 3 our final sample had an average (median) total assets of €319m
(€94m), sales of €196m (€95m) and an EBITDA of €23m (€8m).
Despite the fact that we do lack several companies’ financials, for the ones we have
data, the final sample is not statistically different than the initial sample, in any of the
shown metrics, which seems to indicate that we do have a representative sample.
Table 3 - Descriptive Statistics
This table reports some key financials for the pre transaction year (n-1) for the initial sample and the
final sample. Panel A has the EBITDA, total assets and sales in €m. Panel B shows three key
ratios for our research, investment and EBITDA deflated by beginning of year total assets, and sales
growth. Panel C has some additional characterization ratios. Significance tests for the difference
between samples given by two-tailed Wilcoxon rank sum. We use ***, **, and * to denote
significance at the 1%, 5%, and 10% level (two-sided), respectively. Values are winsorised between
0.05 and 0.95 percentiles for Panels B and C.
PANEL A
Mean
Median
Std. Dev.
N
Dif Initial vs Final (signif.):
PANEL B
Mean
Median
Std. Dev.
N
Dif Initial vs Final (signif.):
PANEL C
Mean
Median
Std. Dev.
N
Dif Initial vs Final (signif.):
EBITDA
Initial
Final
30.0
22.5
6.9
8.2
87.9
51.2
132
92
Total Assets
Initial
Final
366.7
318.8
91.8
94.0
858.3
805.1
167
92
Sales
Initial
Final
267.0
196.4
76.8
94.8
946.9
388.9
150
92
CAPEX / Total Assets
EBITDA / Total Assets
Initial
Final
Initial
Final
0.1027
0.0956
0.1167
0.1199
0.0455
0.0469
0.1013
0.1087
0.1689
0.1449
0.1281
0.1138
123
92
132
92
Sales Growth
Initial
Final
0.1297
0.1314
0.0909
0.0987
0.3275
0.2191
129
92
Fixed Assets / Total Assets
Initial
Final
0.4751
0.4504
0.4362
0.4027
0.3195
0.2862
166
92
22
D/(D+E)
Initial
Final
0.6227
0.5883
0.8419
0.7709
0.3947
0.3974
134
81
Cash / Total Assets
Initial
Final
0.1071
0.1232
0.0380
0.0545
0.1469
0.1549
151
90
3.2 Methodological Considerations
3.2.1 Descriptive evolution
As usual, we compare the evolution of several key metrics from the year before the
transaction (n-1) to the year 1 (n+1), 2 (n+2) and 3 (n+3) after the transaction. The
transaction year is left out as it is difficult to separate what is PE influence.
We analysed 5 different investment or CAPEX intensity measures.
The first two are the year’s CAPEX deflated by the year-end as well as by the year-start
total assets (“Lag total assets” or “Lag TA”).
Some studies do not make a distinction, while others categorically deflate either by
year-start or by year-end total assets. We agree with the view that it makes more sense
to deflate by the year-beginning quantum, as it reflects better the investment decisions
within the year, as the year-end measure is already influenced by the year’s investment
policy. Nevertheless for consistency check purposes we decided to present both.
We also present the CAPEX deflated by the year’s sales as it is customary. By sales we
are actually referring to turnover or operating revenues. Comparability across sector is
enhanced by the use of the latter, as it includes services rendered, which, in some
sectors, can be the only source of income (no actual sales are recorded as per accounting
definitions).
Furthermore, we present 2 additional measures. The measure of “expansionary
CAPEX” deflated also by year-start and year-end total assets.
“Expansionary CAPEX” (sometimes referred to as “growth CAPEX”) envisages being
a measure of how much the company is investing beyond the simple maintenance of its
current production capacity, i.e, in expanding its potential growth. Whilst it is hard to
categorize in each company what is exactly “expansionary” and what is “maintenance”
CAPEX, at least from an outside observer’s perspective, typically one expects the
annual depreciations to be a measure, or at least a proxy, of what a company needs to
invest order to keep its production capacity, as the depreciation itself is calculated, at
least in principle, on the basis of the expected useful life of the equipment.
23
This metric sometimes is just called Net CAPEX and is used as the key metric, but we
believe that it could be useful as support but not as a core measure. The difference
between gross and net CAPEX is D&A, and, as such, can be somehow arbitrary,
impacted by the countries’ fiscal policies, and to some extent manipulated by the
company. Hence, whilst the potential distortions can be acceptable in our expansionary
proxy, a secondary measure, it is our understanding that gross CAPEX captures better
the company’s investment policy.
We define then expansionary CAPEX (“g CAPEX”) as the year’s CAPEX – D&A. One
alternative measure of the expansionary CAPEX is to present CAPEX/D&A. However
this measure is not very stable in companies in early stages as the D&A is still low (e.g.
if the equipments are still in assembly or construction phase are not depreciated) and
inflates the measure when compared to the chosen.
Our definition of CAPEX is then year-end fixed assets minus year-start fixed assets plus
D&A. Three methodological choices arise from this definition.
First, we include the acquisitions effect, following Sousa and Jenkinson (2013) or
Asker et al. (2015) and by opposition of others as Kaplan (1989). While the former
consider acquisitions the latter does not.
The distinction between CAPEX and acquisitions can be important when comparing
public with private firms (hence, in pre-post analysis in public to private transactions)
due to the fact both are alternatives to acquire physical assets (instead of acquiring a
new equipment, the company can acquire another company that has that same
equipment). Nevertheless, private firms are less likely to engage in acquisitions as they
are unable to pay them with stock and, as such, their overall investment is likely to
involve relatively more CAPEX (without acquisitions) than public firms (Asker et al.,
2015).
Second, as it is implicit we also consider investment in intangible assets as CAPEX.
The same basic reasoning as before applies to this choice: the relative importance of
tangible versus intangible assets and CAPEX can be quite different across sectors.
Hence, if we only consider tangible CAPEX, we may fail to acknowledge those
differences.
24
We also analyse the evolution of 5 profitability/cash flow metrics. Earning before
interests depreciation and amortization (“EBITDA”) deflated by year-start and year-end
total assets (commonly referred to as ROA) and by the year’s sales. It is also presented
2 measures of cash flow, defined as EBITDA less CAPEX, deflated both by year-end
and year-start total assets.
We also aimed to analyse the evolution of leverage. For this, instead of using debt to
assets, we opted to use the measures more commonly used on the PE deals financing
structures as covenants, Debt/EBITDA and EBIT/Interest, the latter more accurately a
serviceability measure.
However, when comparing Debt/EBITDA in big samples, where negative values can
occur (and, in fact, we have several), a contradiction arises: a negative value is bad, as it
indicates negative cash flow (proxy) and, hence, no repayment capacity, but the
negative values reduce the overall measure, either mean or median, thus
underestimating it.
A solution could be to consider only positive values, which would, however, still
somehow underestimate the measure of the group’s leverage.
Hence, we decide to present these measures as the difference to the sector median: when
the value is negative it is added (positive sign) to the sector median, when positive is
subtracted to it. Thus, a negative value means a value lower than the sector median and,
albeit somehow still underestimating it, the negative individual values contribute to
increase the group’s overall measure.
3.2.2 Sector Medians
For the 76 4-digit Nace codes in our sample, we extracted all the private companies in
Amadeus database (for the Euro-28 area), which encompassed more than 150 thousand
companies, which, after cleaned of all the missing incomplete data for the whole period,
came down to 17,804 companies. We only kept companies with values for all of our
metrics, during the entire 10 year period, in order to have a fully comparable set of
companies during the decade.
25
3.2.3 Investment Opportunities
A company investment decisions should be influenced by its investment opportunities.
It is not expectable that two similar companies, with different investment opportunities,
have the same investment policy.
Hence, in order to be fully comparable, the investment decisions need to account for the
different investment opportunities. In the empirical investment literature, investment
opportunities are usually represented by Tobin’s Q, which is typically defined as the
firm’s market value to the book value of its assets (as a proxy to its replacement cost).
The problem with this ratio is that is not available for private firms. Asker et al. (2015)
suggests the alternative usage of Sales growth5, referred to as widely used in the
literature, and an “Industry Q”, constructed as a size-weighted average Q of all public
firms in that industry.
A third alternative comes from Campello and Graham (2013) who suggest regressing
public firms’ Q against variables that theoretically explain it 6, and then use the
regression coefficients to generate “Fundamental” Q for each firm, both public and
private7.
We used 4 alternative measures of investment opportunities: industry average Q8, the
industry median Q, the theoretical estimated Q and sales growth.
For industry Q means and medians, we extracted all the public firms for our 76 4-digit
Nace codes in Amadeus, for which we were able to have a balanced panel with market
capitalization, net debt and total assets for the entire period, of 1,836 companies. A
considerable number of public companies missed market capitalization data.
We then calculated sector mean and median Q’s. As we missed or otherwise had a small
sample for some sectors, we considered the values for the 2-digit Nace code,
encompassing 35 sectors.
5
Henceforward, sales growth means sales n / sales n-1, where n is the year where we are estimating
CAPEX or investment intensity (e.g., CAPEX/Total assets).
6
The authors use sales growth, return on assets (EBITDA divided by beginning-of-year total assets), net
income before extraordinary items, book leverage, and year and industry fixed effects.
7
Interestingly, according to the authors, when used in the regression to explain investment, this
Fundamental Q has higher explanatory power than the “real” market Q.
8
We used simple average instead of size weighted average.
26
As detailed in Table 4, for the calculation of predicted Q’s we estimated 5 equations, by
making a regression of the listed companies’ Q with combinations of variables that
theoretically can explain it, as mentioned.
Table 4 - Predicting Tobin's Q
This table reports 5 potentially explaining regressions for Tobin’s Q. Data in Panel with unbalance
data (some missing values mainly leverage). We use ***, **, and * to denote significance at the
1%, 5%, and 10% level (two-sided), respectively. Year and industry (EVCA clusters) fixed effects
(manual dummies, ACM omitted, as automatic cross-section fixed effects would be firm fixed
effects). Year fixed effects test given by redundant fixed effects – Likelihood Ratio.
Heteroskedasticity-consistent standard errors (White diagonal) are shown in italics under the
coefficient estimates. Median sector Q is at the 35 2-digit Nace code level (value per year). Values
were not winsorised.
ROA
(1)
0.2611 *
(2)
0.1434
(3)
0.2624 *
0.1445
(4)
0.3094 ***
0.0481
(5)
0.2670 *
0.1446
0.0074
Lag ROA
0.0147
0.0000
Sales g
0.0000
Leverage (D/TA)
0.0289
0.0589
0.0569
0.0447
Median Sector Q
1.2191 ***
0.063
Constant
0.7061 ***
0.0485
Obs.
Adjusted R-squared
F-statistic
Industry Fixed Effects
Period Fixed Effects
8960
6.4%
41.5 ***
Yes ***
Yes ***
0.7380 ***
0.0471
8632
6.1%
38.2 ***
Yes ***
Yes ***
0.7000 ***
0.0503
8960
6.4%
39.0 ***
Yes ***
Yes ***
0.6930 ***
0.0942
8506
6.5%
36.0 ***
Yes ***
Yes ***
0.01774
0.0476
8959
10.2%
511.7 ***
No
No
Model (5) seems to fit better data and it’s built under the premise that individual Q’s
converge to sector norm and that differences are explained by ROA. Hence, we chose
model (5) to apply to our PE backed companies to estimate a predicted Q for each year.
3.2.4 Investment Regressions
Following the traditional investment literature9 and Asker et al. (2015), we used two
base equations (3.1) which envisages analysing the private equity firms and the peers
9
In the traditional investment literature, or “Q-theory”, investment intensity is usually regressed against
Tobin’s Q or proxy, and other investment determinants, such as ROA. Asker et al. (2015) explains that
that empirical work shows that standard proxies for investment opportunities are not, as neoclassical
theory predicts, a sufficient statistic for investment and that ROA correlates positively with investment.
27
(see below) separately, or together, by adding a PE dummy variable, isolating its effect;
and (3.2) which aims to interact and compare directly the PE firms with its peers,
isolating its impact in each of the explainable variables.
(3.1)
(3.2)
Where I is the investment, or CAPEX, A the beginning of year (or end of year) total
assets, Q the proxy for Tobin’s Q, as discussed above, PE the dummy for private equity
status10 and ROA, meaning the EBITDA divided by total assets (again either beginning
or end of year). The equations include firm (
and
) and year (
and
fixed effects
and allow estimating within-firm variation in investment in response to within-firm
variation in investment opportunities. The PE interaction allows comparing the
investment sensitivities of PE firms with its peers (or public and private firms in the
posterior analysis).
3.2.5 Matching
Not visible, or not controlled, characteristics can determine variations in firms’
investment levels. In order to control or, at least, minimize the effects of any potential
bias, a matching procedure should be implemented. The goal is to compare firms that
are, in fact, similar.
We followed Asker et al. (2015) approach, a calliper-based nearest-neighbour match
adapted to a panel setting, which consists in finding for each public firm, the private
firm, or for each PE backed firm the private non PE backed firm, closest in size in the
same industry code, by requiring the ratio of total assets between the two firms to be
less than 2. In the case there is no match, the observation is discarded. The matches are
held constant in subsequent years to ensure the panel structure remains intact.
Hence, one can include it in the regression, conditioning investment with ROA, a measure of Cash Flow
(as ROA is defined as EBITDA/beginning of year Assets), regardless of its interpretation discussion (as
mentioned in 2.3.2.2. regarding the ICF sensitivity debate).
10
Asker et al. (2015) use a public/private firm dummy.
28
Asker et al. (2015) unlike Bharath et al. (2014) do not match firms based on age and
present an interesting argument against “overmatching” 11. Furthermore, they show that
matched samples result, as one would expect from previous research, in younger, higher
ROA, less cash and more indebted private firms, which nonetheless does not impact the
overall conclusions.
In our matching procedure for the public-private pairs we followed closely Asker et al.
(2015), using 4-digit NACE code instead of 4-digit NAICS. We used VBA coding in
Excel to make the matching computations (please see Appendix II).
As for the PE firms and peer private firms matching we adopted the same methodology,
matching firms on n-1 and keeping them as peers afterwards, relaxing however the <2x
relationship between the firms TA12, as in some industries we ended up with few
companies.
Furthermore, in order to minimize the impact of outliers (as our PE firm sample is
relatively small) we selected not one but two peers (the two closest) and compare to the
mean of the pair.
The percentage of times the ratio between the two firms was >2x was 20% (please bear
in mind that we have 2 sets of pairs which is likely to increase the incidence of
threshold breaches). The median ratio was 1.15, which seems to indicate we have close
peers in size.
We also considered ROA as alternative matching criteria, replacing the total assets with
ROA level13.
11
“The purpose of matching is not to eliminate all observable differences between public and private
firms but to make firms comparable along the dimensions thought to affect the outcome variable of
interest (here: investment). Overmatching on dimensions unrelated to the outcome variable of interest
results in samples that are unrepresentative of their respective populations. In other words, we can make
matched firms arbitrarily similar to each other on arbitrarily many dimensions, but as we do so, the firms
that end up in the matched sample become ever less representative of their respective groups. See
Heckman, LaLonde, and Smith (1999) for an exhaustive discussion of this point.”
12
The relationship is between the max(company A; company B)/min(company A; company B) to force
the results to be always >1; otherwise we would have to work with 0.5 < company A/ company B <2,
which would introduce unnecessary complexity.
13
It seems arguable that it is the dimension, at least isolated, that determines the investment behaviour.
Regardless of the overmatching discussion, we envisaged introducing the profitability (ROA) dimension,
statistically highly correlated with investment intensity. However, in some sectors the 2 dimensions
together determined that we would end up with very different companies in one of the sides or, if we
imposed a maximum relationship (e.g. the two times cap), we would end up without peers. One solution
29
In this case, the incidence of cases where the 2x cap was breached was just 3% (1.01x
median), which seems to indicate that we have closer peers under this criteria. The
mean absolute difference between ROA is 0.23 percentage points, which confirms that
we get really close peers in ROA terms14.
As depicted in Figure 3, we seem to capture relatively similar companies, in terms of
investment policy, by both matching methods, albeit peers matched on ROA seem to be
much more centred and with less outliers/fat tails.
Figure 3 - Kernel Density for CAPEX to TA (left) and Lag TA (right) on year n-1
8
7
7
6
6
PE firms
Peers on TA
Peers on ROA
5
Density
Density
5
4
4
3
3
2
2
1
1
0
0
-.2
-.1
.0
.1
.2
.3
.4
.5
-.2
-.1
.0
.1
.2
.3
CAPEX/TA
.4
.5
.6
.7
CAPEX/ Lagged TA
3.2.6 Endogeneity
One common method to deal with endogeneity is to control for effects that originate the
predisposition. The matched sample procedure tends to minimize this bias. Also the
firm level fixed effects absorbs other non visible within firm differences.
We use both a matching as well as firm fixed effects. As our sample encompasses both
public and private firms pre-deal, we believe this procedure suffices.
could be to relax the sector matching from the 4 digit code to a 1 digit code or to EVCA clusters.
However, we consider that the sector has a considerable influence in investment intensities, playing a
more incisive role than introducing a third, dimension. To illustrate this point, consider glass packaging
companies, a highly CAPEX intensive industry, in which the major investment relates to furnaces
overhauls and that somehow behaves in “waves”, meaning one or two years of sizeable CAPEX followed
by several years of reduced CAPEX, even below what one could consider maintenance CAPEX. Both the
intensity and investment pattern have nothing to do with paper packaging which, under a less
demanding/precise industry classification, would end up both being classified as just “packaging” or even
worse.
14
To be precise as explained in Appendix II, in ROA matching algorithm we did not use a 2x ratio but a
2x the difference between ROAs in order to better deal with negative values.
30
4. PE impact on Investments on Entry – Empirical findings
4.1 Descriptive Evolution
Three years after entry, the median PE backed firm in our sample grew less its sales
than both its sector median (“SM”) and its matched peers in ROA (“PROA”) and in
total assets (“PTA”). As depicted in Figure 4, in sales terms, our PE sample behaved
closely to its PTA and the PROA group was aligned with sector median.
Figure 4 – Median Cumulative Sales (left) and Total Assets (right) Growth
This Figure depicts the sales and total assets median cumulative evolution for the 3 years after the
buyout when compared to the year before the buyout. Below the charts we show the numbers.
Significance tests given by the two-tailed Wilcoxon signed rank test for the cumulative change from
year n-1 to n+i and the two-tailed Wilcoxon rank sum for the difference between group. We use ***,
**, and * to denote significance at the 1%, 5%, and 10% level (two-sided), respectively. Values are
winsorised between 0.05 and 0.95 percentiles.
Cumulative Sales growth
PE
Peer
ROA
Peer
TA
Sector
Median
Cumulative Total Assets growth
30%
30%
25%
25%
20%
20%
15%
15%
10%
10%
5%
5%
0%
0%
1
PE
Peer ROA
Peer TA
SM
Differences:
PE – Peer ROA
PE – Peer TA
PE – Sector Med
Peer TA – Sector Med
Peer ROA – Sector Med
n+1
0.1030**
0.1125***
0.0803***
0.1131***
2
n+2
0.0213*
0.1481***
0.0446***
0.1131***
3
1
years
n+3
0.0389
0.1983***
0.0554***
0.1528***
2
3
Years
n+1
0.0769**
0.0535***
0.0603***
0.1467***
n+2
0.1089***
0.1480***
0.0866***
0.2171***
n+3
0.1903***
0.1546***
0.1058***
0.2657***
**
***
***
**
***
**
***
**
***
*
**
**
**
At year 3, the difference was statistically significant (two-sample Wilcoxon rank-sum Mann-Whitney – test) at 5% and 1% for the sector median and PROA. Furthermore,
PTA’s cumulative sales growth at year 3 was also statistically different (5%) from
sector median while PROA was not.
31
However, in terms of total assets the evolution was quite different, with our PE firms
differing from sector median during the first 2 years, but not in the third, and never
differing statistically from its matched peers.
The peers, however, had a statistically different evolution from the SM (1% significance
for PTA and 5% for PROA).
Regarding investment intensity, Table 5 shows in Panel A the mean and median
percentage points change in the 5 alternative metrics, from year n-1 to the first three
years after the entry, and for the 4 groups (PE, SM, PTA and PROA). The value of
change is given by (Metricn+i – Metricn-1) x 100, where Metric is one of the 5 alternative
investment intensity ratios, for each group, i is the year after entry, from 1 to 3.
Data shows that by year 3 there is a statistically significant reduction, under all of our
selected metrics. Despite the fact that the averages are considerably higher than
medians, as a result of considerable amount of extreme higher values, the same
conclusion is derived from both statistics.
For example, the average (median) PE backed firm by n+3 had reduced by 4.048
(1.020) percentage points (“pp”) its CAPEX deflated by its beginning of year total
assets (Lag TA). This implies that the mean in year 3 (not shown) is shy from half of the
one in n-1.
Furthermore, the expansionary CAPEX intensity proxy (g CAPEX / Lag TA) reduction
implies that the median PE backed firm cut its expansionary CAPEX to virtually 0 in
n+3.
Also worth highlighting that some statistically significant evolutions by n+1 stop being
so by year n+2, which seems to be a year a major volatility and where apparently some
companies partially correct the CAPEX cuts performed during the previous year.
Putting this evolution into context, the sector medians also present a clear reduction
trend in investment intensity. By year n+3, only the median change in CAPEX / Sales is
not statistically significant. The same applies to PTA.
32
Table 5 - Investment Intensity change after entry
This table reports in Panel A the mean and median percentage points change in 5 investment intensity metrics for the first 3 years after entry when compared to the predeal year and in Panel B the difference between medians and medians for the groups. The values are presented for the PE firms, the sector median (of each PE firm) and
the peers matched on total assets and on ROA. Significance tests given by the t-statistic for means and two-tailed Wilcoxon signed rank test for medians in Panel A and
Anova F-Test and the two-tailed Wilcoxon rank sum for means and medians, respectively, in Panel B. We use ***, **, and * to denote significance at the 1%, 5%, and
10% level (two-sided), respectively. Values are winsorised between 0.05 and 0.95 percentiles.
Panel A
n-1 to n+1
CAPEX / TA
Mean
Median
Std. Dev.
CAPEX /Lag
Mean
TA
Median
Std. Dev.
G CAPEX / TA Mean
Median
Std. Dev.
G CAPEX / Lag Mean
TA
Median
Std. Dev.
CAPEX / Sales Mean
Median
Std. Dev.
-2.434
-1.203
11.789
-3.428
-0.384
14.761
-2.413
-1.084
11.624
-2.992
-1.142
13.529
-2.434
-1.203
11.789
*
**
*
**
**
*
*
Panel B
Mean
Median
CAPEX /Lag
Mean
TA
Median
G CAPEX / TA Mean
Median
G CAPEX / Lag Mean
TA
Median
CAPEX / Sales Mean
Median
3.101
0.920
4.753
0.476
2.881
1.132
4.314
1.326
9.353
1.077
n-1 to n+3
-1.494
-0.695
11.069
-2.735 *
-0.271
15.037
-1.665
-0.433
10.660
-2.343
-0.458
13.857
-1.494
-0.695
11.069
-2.588
-1.482
10.131
-4.048
-1.020
14.180
-2.570
-0.873
9.633
-3.749
-1.019
12.604
-2.588
-1.482
10.131
**
**
***
**
**
**
***
**
**
**
PE Firms minus Sector Median
n+1
n+2
n-1
CAPEX / TA
PE Firms
n-1 to n+2
***
***
***
***
***
1.584
0.362
2.490 **
0.190
1.045
0.239
1.951 *
0.285
8.678 ***
-0.192
2.325
0.742
2.987
0.439
1.573
0.377
2.403
0.449
9.868
1.681
**
**
***
***
n-1 to n+1
-0.918
-0.528
2.001
-1.165
-0.737
2.209
-0.576
-0.220
1.541
-0.629
-0.244
1.641
-0.665
-0.374
1.784
***
***
***
***
***
***
***
***
***
***
n+3
1.262
0.581
1.687
0.702
0.676
0.162
0.964
0.180
5.738
1.056
Sector Median
n-1 to n+2
-0.718
-0.598
1.582
-0.969
-0.821
1.910
-0.357
-0.119
1.244
-0.432
-0.137
1.351
-0.238
-0.305
1.853
***
***
***
***
***
**
***
***
***
n-1
**
**
*
***
n-1 to n+3
-0.750
-0.631
1.717
-0.982
-0.756
2.008
-0.365
-0.189
1.190
-0.399
-0.269
1.280
-0.036
-0.179
2.482
***
***
***
***
***
***
***
***
*
Peers matched on TA
n-1 to n+1
n-1 to n+2
n-1 to n+3
-1.285
-0.771
9.144
-1.939 *
-1.143
10.946
-1.129
-0.708
8.558
-1.629
-0.254
10.303
-9.153 *
-0.981
46.065
PE Firms minus Peers on TA
n+1
n+2
0.454
-1.002
0.658
-1.502
0.899
-0.436
0.885
-0.295
-10.066 *
-2.409
-0.695
-0.595
-0.832
-0.782
-0.385
-0.173
-0.477
-0.430
-2.254
-2.616
33
0.073
-0.462
-0.218
-0.773
0.241
0.514
0.111
0.378
-2.522
-1.495
-1.114
-1.132
8.746
-1.859 *
-1.932 *
10.356
-1.007
-0.346
8.421
-1.568
-0.843
10.215
-7.267
-0.825
44.418
-2.720
-1.868
9.499
-3.716
-1.907
11.358
-2.388
-1.239
9.203
-2.954
-1.312
10.883
-12.892
-1.664
47.363
***
***
***
***
**
**
**
**
**
**
n+3
n-1
0.585
0.301
0.326
0.436
0.717
-0.091
0.090
-0.146
-0.824
-0.384
0.685
-1.003
1.158
-2.024
0.948
0.530
1.259
0.667
-1.772
-0.557
Peers matched on ROA
n-1 to n+1
n-1 to n+2
n-1 to n+3
-1.805
-1.356
8.597
-1.780
-1.651
11.599
-1.909
-0.844
9.134
-1.855
-0.966
11.301
-5.960
-0.966
36.738
**
**
***
**
*
*
0.253
-0.552
8.840
3.061
-0.422
18.340
0.787
-0.319
9.367
3.412 *
-0.076
17.462
4.659
-0.225
41.594
PE firms minus Peer on ROA
n+1
n+2
0.056
0.199
-0.490
-0.524
0.444
0.789
0.122
0.813
2.847
-0.389
-1.062
-0.027
-4.638 **
-0.704
-1.504
0.100
-4.496 **
-0.083
-6.155
-0.711
-0.473
-0.656
8.806
-0.662
-1.274
12.513
-0.181
-0.673
8.848
0.003
-0.862
12.125
5.180
-0.107
38.166
n+3
-1.430
0.158
-2.227
-0.002
-1.440
-0.653
-2.493
-0.537
-10.603
-0.229
**
*
**
**
A generalized reduction in investment is not surprising, as our sample is clearly
concentrated in the financial crisis period.
Table 5, Panel B shows, for each year, from pre-entry (n-1) to year 3 after buyout (n+3),
the percentage points difference between the PE investment intensity metric, and each
of the other peer groups, i.e., PE vs. SM, PE vs. PTA and PE vs. PROA. The value is
then (MetricPEi – MetricPGi) x100, where PEi and PGi are the value of the metric for the
PE and for the peer group, respectively, in year i, the relative year from the deal, from 1 to 3.
When comparing the values between the groups, the conclusions seem different. In year
n-1 PE firms invest statistically significant above the sector median, and, while this
higher investment still holds in year n+3, it seems that PE firms somehow converged to
sector norm, by reducing much more its investment levels.
Comparison with PTA is not as clear cut, as differences are not consistent and not
statistically different. Data does not support a difference in behaviour between PE and
PTA firms.
Peers matched on ROA, show a very dissimilar picture. Whilst no statistical significant
difference is found in year n-1, by year n+3 all mean investment metrics, with exception
of CAPEX/TA, show PE firms investing less than its peers (medians also show the
same sign but the difference is not statistically significant). For example, mean CAPEX
/Lag TA is 2.5 pp lower in PE than in PROA firms and mean CAPEX/Sales is 10.6 pp
lower.
The difference in behaviour is especially significant in our expansionary CAPEX
measure. We do not show the absolute (metric/group/year) values, but for illustrative
purposes, we can indicate that whilst a PE backed firm pre buyout invested on average
(median) 4% (1.7%) of its beginning of year assets, the peers (on ROA) invested 3.6%
(0.7%). By the end of the 3rd year, while the behaviour remained virtually unchanged
for the peer group, the PE backed firms had cut its expansionary CAPEX to 1.1% on
average, and the median firm was only replacing assets (0% expansionary CAPEX).
34
Regarding profitability/cash-flow evolution, as show Table 6, Panel A, similar to what
happened with CAPEX, data shows that sector medians statistically (mostly significant
at 1%) reduce over the period, and that the decline is progressive. The exception is
EBITDA / Sales, which peak reduction occurs in year n+2 and then recovers, to the
point that in n+3 the reduction is not statistically significant.
Contrasting with this trend, PE backed firms improve, as found in most of the literature,
their profitability levels by year n+1 the improvements are statistically significant (at
5% and 1%) with exception of ROA calculated on beginning of year TA (ROA Lag).
However, unlike the literature, we somehow witness a reversal, to the point that by year
n+2 and n+3 the improvements disappear and even represent a reduction, albeit not
statistically significant. The exception is cash flow to lagged TA, that in n+3 still
represents a significant (at 10%) 3.2 pp improvement, no doubt due to the reduction in
CAPEX.
In relation to our control groups, PTA also shows reductions in profitability levels,
some of which statistically significant. For example, in year n+3, ROA calculated on
lagged TA evidences a reduction of -1.5 and -0.7 pp in mean and median, respectively
(5% and 10% significance).
As PE firms, PTA firm’s cash-flow also increases, although with significance only in
mean terms and when deflated by lagged TA.
As for PROA firms, both ROA and ROA Lag show significant reductions of c. -2.5 and
-1.8 pp in mean/median terms. Unlike the peer group matched on TA and PE firms,
PROA group also reduces its cash-flows, significant against lagged TA.
Comparing the groups, as shown in Table 6, Panel B, for each year, we can see that PE
firms increase the distance for sector medians in profitability. Cash flow measures were
lower than sector norm in n-1, albeit not significant, and become higher and statistically
different by n+3.
35
Table 6 - Profitability and Cash-Flow change after entry
This table reports in Panel A the mean and median percentage points change in 5 profitability / cash flow metrics for the first 3 years after entry when compared to the predeal year and in Panel B the difference between medians and medians for the groups. The values are presented for the PE firms, the sector median (of each PE firm) and
the peers matched on total assets and on ROA. Significance tests given by the t-statistic for means and two-tailed Wilcoxon signed rank test for medians in Panel A and
Anova F-Test and the two-tailed Wilcoxon rank sum for means and medians, respectively, in Panel B. We use ***, **, and * to denote significance at the 1%, 5%, and
10% level (two-sided), respectively. Values are winsorised between 0.05 and 0.95 percentiles.
Panel A
n-1 to n+1
ROA
Mean
Median
Std. Dev.
ROA Lag
Mean
Median
Std. Dev.
EBITDA / Sales Mean
Median
Std. Dev.
Cash Flow / TA Mean
Median
Std. Dev.
Cash Flow / Lag Mean
TA
Median
Std. Dev.
2.743
1.571
10.507
1.512
0.299
14.763
2.770
0.900
13.056
5.745
3.507
16.135
6.269
3.940
18.926
**
**
**
**
***
***
***
***
Panel B
Mean
Median
ROA Lag
Mean
Median
EBITDA / Sales Mean
Median
Cash Flow / TA Mean
Median
Cash Flow / Lag Mean
TA
Median
2.073
1.384
3.829
1.087
3.859
2.799
-0.155
-0.190
-0.851
-0.976
n-1 to n+3
-0.937
-0.729
11.207
-2.341
-0.716
14.609
-0.824
-0.313
15.672
0.628
0.401
14.858
1.083
-0.110
18.297
-0.853
-1.050
10.587
-1.880
-1.514
14.118
1.558
0.743
16.109
1.875
0.700
14.037
3.235 *
0.842
17.613
PE Firms minus Sector Median
n+1
n+2
n-1
ROA
PE Firms
n-1 to n+2
*
**
**
*
5.415
2.768
6.187
2.208
6.818
5.139
5.772
2.775
5.736
2.703
***
*
***
***
***
***
***
***
***
2.192
1.840
2.991 **
2.231
3.467 *
1.949
1.139
0.443
1.110
1.184
n-1 to n+1
-0.599
-0.320
1.904
-0.846
-0.804
1.927
-0.189
-0.083
2.180
-0.182
-0.034
2.453
-0.317
-0.187
2.641
***
***
***
***
n+3
2.551
2.131
3.741
2.131
5.605
2.123
2.727
0.892
3.598
1.314
Sector Median
n-1 to n+2
-1.055
-1.106
1.935
-1.503
-1.354
2.231
-0.431
-0.526
1.983
-0.666
-1.091
2.223
-0.878
-1.360
2.248
***
***
***
***
**
***
***
***
***
***
n-1
**
***
***
**
***
n-1 to n+3
-1.331
-1.335
1.763
-1.793
-1.872
2.184
-0.187
-0.564
2.336
-1.007
-1.230
2.005
-1.214
-1.330
2.020
***
***
***
***
**
***
***
***
***
Peers matched on TA
n-1 to n+1
n-1 to n+2
n-1 to n+3
-0.163
-0.129
5.162
-0.491
0.090
6.465
-4.474
-0.147
28.412
0.792
0.222
10.084
1.562
1.379
12.467
PE Firms minus Peers on TA
n+1
n+2
1.540
1.060
3.187 *
0.368
-8.632 **
-0.549
1.218
0.410
2.196
1.297
4.446
1.827
5.190
1.246
-1.389
2.997
6.171
3.375
6.903
3.853
36
***
***
***
***
***
***
1.920
0.460
2.558
0.367
-1.019
-0.863
2.472
0.084
3.203
1.049
-1.316
-1.336
5.704
-1.712
-1.400
6.785
-8.437
-2.234
30.635
-0.627
-0.647
10.163
0.076
0.212
11.614
**
***
**
***
***
**
n+3
1.479
0.796
2.846 *
1.421
-2.907
-0.009
1.812
1.341
3.110 *
1.286
-0.792
-0.929
6.381
-1.539
-0.756
7.400
-4.167
-1.634
30.963
1.281
0.803
10.849
2.320
0.966
12.484
*
**
*
*
*
n-1
-0.419
-0.135
1.644
-0.679
-0.624
0.872
-1.015
0.543
-0.832
0.108
Peers matched on ROA
n-1 to n+1
n-1 to n+2
n-1 to n+3
-1.341
-0.265
8.393
-1.286
-1.267
8.696
0.117
-0.270
24.969
1.588
0.846
14.483
0.409
-1.068
14.958
-2.132
-1.290
9.291
-2.002
-1.262
9.898
1.844
0.028
15.892
-3.058
-1.105
15.361
-5.584
-1.594
21.187
**
**
*
**
*
**
*
PE firms minus Peer on ROA
n+1
n+2
3.664
1.075
4.443
0.562
2.029
3.593
3.142
0.787
5.028
0.858
**
**
*
**
0.776
1.001
1.304
1.495
-3.292
-0.817
2.671
0.710
5.836 **
1.470
-2.503
-1.783
8.365
-2.589
-1.850
9.189
1.723
-1.685
21.845
-2.101
-0.724
12.407
-2.476
-1.048
14.147
***
***
***
***
*
n+3
1.230
1.029
2.353
0.162
-0.789
1.098
2.962 *
1.987
4.879 ***
2.104
In relation to PTA group, PE firms considerably increase the difference in the first year
(n+1), although only means are statistically different in ROA terms (e.g. 5.2 pp higher
in PE than PTA in mean ROA Lag). In cash flows, both means and medians are
statistically different. However, given the aforementioned reversal in PE firms’
performance, the differences are no longer significant by year n+3, with exception of
mean ROA Lag and mean Cash-Flow/Lag TA.
As it’s not surprising there are no statistical differences between means/medians, in n-1,
in PE vs PROA, as the latter was matched exactly on ROA. By year n+3 however, cashflow measures are significantly higher in PE firms, due to the mentioned cutbacks in
CAPEX, whilst PROA firms basically kept their investment levels.
Finally, in relation to our leverage and serviceability metrics, the results, as shown in
Table 7, Panel A, show some evidence of increased leverage in our PE firms, and
reduced interest cover (IC) in our peer groups.
Table 7 - Leverage and Interest Cover
This table reports in Panel A the mean and median change in the difference of each group’s debt to
EBITDA and EBIT to interest (IC) to the sector median for the first 3 years after entry when compared to
the pre-deal year and in Panel B the difference between medians and medians for the groups. The values
are presented for the PE firms and the peers matched on total assets and on ROA. Significance tests given
by the t-statistic for means and two-tailed Wilcoxon signed rank test for medians in Panel A and Anova
F-Test and the two-tailed Wilcoxon rank sum for means and medians, respectively, in Panel B. We use
***, **, and * to denote significance at the 1%, 5%, and 10% level (two-sided), respectively. Values are
winsorised between 0.05 and 0.95 percentiles.
Panel A
PE minus SM
PTA minus SM
PROA minus SM
n-1 to n+1 n-1 to n+2 n-1 to n+3 n-1 to n+1 n-1 to n+2 n-1 to n+3 n-1 to n+1 n-1 to n+2 n-1 to n+3
Debt/
Mean
-0.8
1.3
1.9
-1.3 *
-1.1
0.1
-0.1
0.6
0.3
EBITDA
Median
-0.2
0.2 *
0.2 *
0.0
-0.1
-0.1
-0.1
0.1
0.0
Std. Dev. 10.0
9.9
10.5
6.6
6.6
8.8
5.0
6.7
5.8
IC
Mean
3.6
-5.3
29.4
-53.3 *
-82.1 **
-73.7 ** -86.1 **
-83.2 **
-98.6 ***
Median
0.3
0.1
0.4
-0.5
0.3
0.6
-0.7
-0.4
-1.2
Std. Dev. 151.2
158.9
232.3
268.8
341.5
322.5
347.0
337.4
341.7
PE minus SM vs. PTA minus SM
PE minus SM vs. PROA minus SM
Panel B
n-1
n+1
n+2
n+3
n-1
n+1
n+2
n+3
Debt/
Mean
0.0
0.6
2.4 *
1.9
1.4
0.7
2.1
3.0 **
EBITDA
Median
-0.7 **
-1.1
-0.4
0.2
-0.2
0.2
0.1
-0.3
IC
Mean
-67.4
-17.9
6.5
32.0
-94.2 *
-11.8
-19.2
30.1 *
Median
0.8
0.3
0.2
0.8
-4.1
-2.8
-3.4
-1.6
37
Comparing the groups (Table 7, Panel B), results show that leverage increases in
relation to the PROA group and also to PTA, although the latter without statistical
significance. For example, the mean difference in debt to EBITDA towards the sector
median is 3.0 higher in n+3 in PE firms than in PROA group (5.8x vs. 2.8x – not shown
in the tables, which only shows the difference of 3.0). Medians have contradictory
signs, but are not statistically significant.
Interestingly, interest cover is also higher in PE backed firms than in PROA peers, in
year n+3, which could be explained by lower interests, on the back of PE higher
negotiating power and broader access to financing sources.
4.2 Measuring Investment Opportunities
Although the empirical research shows that other factors explain differences in
investment intensity, namely and remarkably ROA, the according to the neoclassical
approach those differences should simply be a result of different investment
opportunities (Asker et al., 2015).
We arranged our data in panel1 and estimated for our PE firms sample the sensitivity of
the investment intensity to investment opportunities proxies and profitability/cash flow
measure, as specified in equation (3.1).
In Table 8, we show in the columns under “Proxies to Investment Opportunities”, for
our main investment intensity metric, CAPEX/Lag TA, the estimation of the equation
with 4 alternative proxies to Tobin’s Q: (1) sales growth; (2) estimated or “theoretical
Q”; (3) median sector Q and; (4) mean sector Q. For the other investment intensity
metrics we show under “Alternative Investment Intensity Measures” (regressions 5 to 8)
the estimation, using only the Sales growth proxy to Tobin’s Q.
1
We include the civil year as the year identifier, which results in an unbalanced panel, as our data is
balanced relatively (to deal) year terms. We consider, nonetheless, the advantaged of fixing civil years
effect, for example, to isolate crisis years. An alternative “crisis” dummy did not seem to be a good
approach (we did estimate several alternative formulations), as our sample encompasses very different
European countries, which as commonly known, were affected by the crisis in different years (the impact
in Europe was not simultaneous), and thus, identifying the years which could fall under the “crisis”
dummy could be tricky.
38
The results show that as reported in the literature, sales growth seems to hold as a good
investment opportunities proxy, given its statistical significance across the several
investment intensities measures we defined.
Table 8 - PE firms sensitivity to Investment Opportunities
This table reports the estimation of Equation (3.1) for PE firms, testing the different proxies to investment
opportunities as discussed, sales growth (1), Estimated Q (2), median sector Q (3) and mean sector Q (4).
Investment opportunities proxies are tested against CAPEX to lagged total assets. We also present
alternative measures of investment Intensity, regressed against sales growth as investment opportunities
proxy. All regressions include company and year fixed effects. Fixed effects test given by Redundant
Fixed Effects – Likelihood Ratio. Heteroskedasticity-consistent standard errors (White diagonal) are
shown in italics under the coefficient estimates. We use ***, **, and * to denote significance at the 1%,
5%, and 10% level (two-sided), respectively. Values are winsorised between 0.05 and 0.95 percentiles,
with exception of estimated, median and mean Q’s.
Proxies to Investment Opportunities
CAPEX / Lag TA
(1)
(2)
(3)
(4)
ROA
Alternative Investment Intensity Measures
CAPEX / CAPEX / G CAPEX G CAPEX /
TA
Sales
/ TA
Lag TA
(5)
(6)
(7)
(8)
0.0528
0.0206
0.0558
ROA Lag
0.2163 *** 0.2662 *** 0.2704 *** 0.2720 ***
0.058
Sales growth
0.059
0.059
0.059
0.0580 ***
0.1651 ***
0.147
0.0657 *** 0.1030 *
0.021
Estimated Q
0.0537
0.2011
0.019
0.055
0.056
0.0523 *** 0.0455 **
0.018
0.020
0.0461 *** 0.0834 *** 0.0091
-0.0043
0.0071
0.018
Median Sector Q
0.0199
0.021
Mean Sector Q
0.0154
0.011
Constant
0.0310 *** 0.0195
0.009
Company /Year
Fixed Effects:
Obs.
Adjusted R2
F-statistic
Yes ***
0.021
Yes ***
0.0096
0.0090
0.021
0.015
Yes ***
Yes ***
460
460
460
460
31.5%
29.4%
29.5%
29.8%
3.09 ***
2.89 ***
2.90 ***
2.92 ***
0.008
Yes ***
0.022
Yes ***
0.007
Yes ***
0.008
Yes ***
460
460
460
460
27.7%
38.9%
15.5%
18.0%
2.74 ***
3.89 *** 1.83 ***
2.00 ***
Our alternative measures, such as sector’s median or mean Q or even our estimate of the
“theoretical” Q don’t seem to hold. We recall that our Estimated Q was calculated with
model (5) in Table 4.
As reported in the investment literature, our sample of PE firms CAPEX intensity seems
to be explained by variations in profitability/cash flow and investment opportunities,
given by the Tobin’s Q proxy (sales growth in our case).
39
4.3 Conditional Investment Intensities – PE firms vs. Peers
Our research question, however, is not determining the explainable variables for PE
firms per se, but to determine how they compare with their peers, after controlling for
those same explanatory variables.
Equation (3.1) adapted with a PE dummy variable allows us to isolate the PE impact
and equation (3.2) allows us to explore this through the interaction of the firm status
(dummy variable) with the explainable variables and thus measure the impact of the
specific nature of PE backed firms.
We have estimated the equations for all the alternative investment intensity measures,
we’ve considered previously, and for both set of peers at the same time as well as
separately.
We also separate the “pre PE” period (n-1 or n-1 and n if necessary) from the “post PE
entry” period (n+1 to n+3) in order to assess the impact of the change in ownership.
Several conclusions can be drawn from the results, which are shown in Table 9.
First, Equation (3.1) as shown in Table 9 Panel A, seems to provide little evidence of
any impact of the PE intervention in firms’ investment intensity, controlling for
investment opportunities and cash-flow/profitability. In fact, only regression (4) in
Table 9 Panel A, referring to the post PE entry and in a sample of PE and peers matched
on ROA and regarding CAPEX / Lag TA, shows such an impact. In this model, the PE
firm, everything else constant, has a CAPEX/Lag TA of lower in 1.28 pp (sig. 10%) in
relation to its matched peers.
Second, with exception CAPEX/Sales in PROA and PTA samples only (regressions
(51) and (54) in Table 9 Panel A), all our regressions show sales growth, our proxy to
investment opportunities, as statistical significant, and the majority of times at 1%.
Third, and more important, Equation (3.2), as estimated and shown in Panel B, which
allows us to develop and further build the results from equation (3.1) - it enables us to
separate the explanatory variables by firm type -, seems to provide some evidence of a
negative impact of PE intervention in investment.
40
Table 9 - Sensitivity to Investment Opportunities across PE and matched peers
This table reports the estimation of, in Panel A Equation (3.1) adapted with a PE dummy, and in Panel B Equation (3.2) which allows the analysis of within-firm
variation to differences in the sensitivity of investment intensity to investment opportunities (sales growth as proxy), and profitability, between PE firms and other
private firms considered as the best “comparable” according to our two matching criteria. Each of the 5 Investment intensity measure is regressed in three ways (i) all
firms, i.e., PE firms and peers matched by both methods (column “All” under each metric);(ii) PE firms and peers matched by ROA (column “PE + PROA”) and; (iii)
PE firms and peers matched on TA (column “PE+PTA”). For each sample two regressions are estimated to separate the impact of the PE entry: n-1 (Panel B needs to
be n-1 to 0 in pre PE period to allow Company FE) vs. n+1 to n+3 periods. Cross section and year fixed effects (FE) are included. In Panel A the specification only
allows for sector (4-digit Nace) FE, whilst in Panel B we are allowed to include company FE. The majority of equations have significant FE (Redundant Fixed Effects
– Likelihood Ratio) - kept in all for consistency. Heteroskedasticity-consistent standard errors (White diagonal) are shown in italics under the coefficient estimates. We
use ***, **, and * to denote significance at the 1%, 5%, and 10% level (two-sided), respectively. Values are winsorised between 0.05 and 0.95 percentiles.
CAPEX / Lag TA
All
PANEL A
Period
ROA Lag
(1)
n= -1
0.1560 *
Sales g
0.1949 ***
0.084
PE dummy
0.020
0.050
-0.0098
-0.0050
0.0952 **
2
18.4%
1.8 ***
(13)
n= -1 to 0
0.0545
0.134
0.129
0.0518
0.046
0.029
x PE
0.0658
-0.0556
0.061
0.012
0.3937 **
0.155
0.0970 ***
0.040
0.0492 ***
0.010
0.150
-0.0376
0.189
-0.0332
0.014
0.174
0.4470 **
0.193
0.1245 ***
0.056
0.038
0.1527 **
-0.0858 *
0.070
0.0731 ***
0.012
0.048
0.0466 ***
0.012
0.062
0.0792 ***
0.083
0.2442 ***
0.042
0.0633 ***
0.040
0.019
0.050
0.024
0.051
0.020
-0.0059
0.001
-0.011
-0.0104
-0.0033
0.012
10.4%
1.8 ***
(16)
n= 1 to 3
-0.0760
0.152
0.1056 **
PE + PTA
(11)
(12)
n= -1
n= 1 to 3
0.0055
0.0201
-0.0049
0.0203
0.069
0.041
0.0829 ***
PE + PROA
(9)
(10)
n= -1
n= 1 to 3
0.2649 *
0.0291
0.023
0.008
27.9%
1.9 ***
(15)
n= -1 to 0
0.0224
0.089
0.1909 ***
(8)
n= 1 to 3
0.0169
-0.0077
0.0103
0.011
(7)
n= -1
-0.0007
0.053
0.012
9.0%
2.0 ***
(14)
n= 1 to 3
-0.0071
0.047
0.0739 ***
All
-0.0174
0.0793
Sales g
0.0648 ***
0.025
-0.0128 *
0.082
0.2715 ***
0.006
-0.0686
0.181
0.064
0.0915 ***
Expansionary CAPEX / Lag TA
PE + PTA
(5)
(6)
n= -1
n= 1 to 3
0.1318
0.1163 **
0.0168
x PE
Constant
0.149
0.1117 **
0.041
0.048
Adjusted R
F-statistic
PANEL B
Period
ROA Lag
0.044
0.0921 ***
-0.0107
0.010
Constant
(2)
n= 1 to 3
0.1033 **
PE + PROA
(3)
(4)
n= -1
n= 1 to 3
0.3284 **
0.0545
0.059
30.0%
2.0 ***
(17)
n= -1 to 0
0.1424
0.007
0.010
0.0279 **
0.0987 **
0.012
10.2%
1.7 ***
(18)
n= 1 to 3
0.0782
0.048
8.7%
1.3 *
(19)
n= -1 to 0
0.0633
0.195
0.194
0.139
-0.1518
0.3003
-0.1437
0.220
0.1379 ***
0.210
0.178
0.045
0.0468
0.006
0.012
0.008
0.011
0.006
0.0040
0.0776
0.0031
0.0255
0.0120
0.010
5.9%
1.6 ***
(20)
n= 1 to 3
-0.0217
0.123
0.3524 **
0.152
0.0892 *
0.057
0.042
0.042
0.028
-0.0280
-0.008
0.0564
-0.0564
0.068
0.0539 ***
0.014
2
30.4%
23.0%
36.4%
24.9%
28.5%
Adjusted R
F-statistic
1.8
1.9 ***
2.1 ***
1.9 ***
1.8 ***
Obs.
552
828
368
552
368
Sector (Panel A) / Company (Panel B) and Year Fixed Effects Included in all equations
0.049
0.0331 ***
0.011
27.1%
2.1 ***
552
41
0.060
0.0220 *
0.013
19.7%
1.5 ***
552
0.039
0.0091
0.010
13.4%
1.5 ***
828
0.065
0.013
0.061
20.1%
1.6 **
(21)
n= -1 to 0
0.0154
5.1%
1.4 **
(22)
n= 1 to 3
-0.0827
18.6%
1.5 **
(23)
n= -1 to 0
0.2067
0.153
-0.0998
0.189
-0.0367
0.173
0.3954 **
0.194
0.1176 ***
0.051
0.038
0.1437 **
-0.0852 *
0.068
0.0339 ***
0.013
26.2%
1.7 ***
368
0.012
8.0%
1.6 ***
(24)
n= 1 to 3
0.0514
0.196
0.179
-0.2760
0.2574
0.219
0.1178 **
0.196
0.0397
0.053
0.038
-0.0252
-0.0154
0.048
0.066
0.045
0.0079
0.0102
-0.0049
0.012
13.7%
1.5 ***
552
0.013
15.5%
1.4 **
368
0.011
18.0%
1.6 ***
552
Table 9 (Continued)
Two alternative Investment intensity measures CAPEX and Expansionary CAPEX deflated with end-of-year rather than beginning-of-year Total Assets.
CAPEX / TA
All
PANEL A
Period
ROA
(25)
n= -1
-0.1134 *
0.067
Sales g
0.1498 ***
PE dummy
Adjusted R
F-statistic
PANEL B
Period
ROA
0.055
0.0766 ***
0.078
0.2068 ***
0.042
0.0584 ***
(31)
n= -1
0.0605
0.069
0.1487 ***
(32)
n= 1 to 3
0.0247
0.038
0.0830 ***
PE + PROA
(33)
(34)
n= -1
n= 1 to 3
0.0692
-0.0237
0.183
0.0971 ***
0.055
0.0842 ***
PE + PTA
(35)
(36)
n= -1
n= 1 to 3
0.0827
0.0466
0.076
0.2188 ***
0.045
0.0644 ***
0.015
0.035
0.019
0.039
0.018
0.031
0.016
0.035
0.019
0.042
0.019
-0.0001
0.0022
-0.0027
-0.0035
0.0018
-0.0053
-0.0047
-0.0030
-0.0057
-0.0091
-0.0033
0.005
0.010
0.006
0.009
0.005
0.0018
0.0925
-0.0006
0.0258
0.0083
0.040
0.009
0.057
0.011
0.050
9.9%
1.4 **
(37)
n= -1 to 0
-0.2277 *
7.5%
1.8 ***
(38)
n= 1 to 3
-0.2794 **
10.1%
1.3
(39)
n= -1 to 0
-0.2369 **
6.0%
1.4 **
(40)
n= 1 to 3
-0.3874 **
20.0%
1.6 **
(41)
n= -1 to 0
-0.0528
0.125
0.111
x PE
-0.0565
Sales g
0.0470
0.032
0.022
x PE
0.0490
-0.0479
0.192
0.049
Constant
0.176
0.0995 ***
All
0.029
0.0919 **
2
0.036
0.0781 ***
Expansionary CAPEX / TA
PE + PTA
(29)
(30)
n= -1
n= 1 to 3
-0.0707
-0.0421
-0.0004
0.008
Constant
(26)
n= 1 to 3
-0.0599 *
PE + PROA
(27)
(28)
n= -1
n= 1 to 3
-0.0095
-0.0437
0.0399 ***
0.011
0.4366 ***
0.141
0.0955 ***
0.032
0.0237 ***
0.009
0.119
-0.0024
0.175
-0.0002
0.159
0.5358 ***
0.178
0.1228 ***
0.037
0.029
0.0924 *
-0.0759 *
0.052
0.0436 ***
0.011
0.225
-0.2352
0.269
0.1052 **
0.011
9.0%
1.6 ***
(42)
n= 1 to 3
-0.1831
0.156
0.3471 *
0.008
0.0861 **
0.040
17.0%
1.7 ***
(43)
n= -1 to 0
-0.2120 *
0.114
0.0365
0.178
0.192
0.0414
0.0453
0.005
0.010
0.006
0.010
0.0152
0.0881
0.0078
0.0158
0.009
9.1%
2.0 ***
(44)
n= 1 to 3
-0.2739 **
0.112
0.4460 ***
0.138
0.0965 ***
0.044
0.030
0.035
0.022
-0.0091
-0.0004
0.0664
-0.0391
0.037
0.057
0.0233 **
0.0280 *
0.011
0.014
2
19.0%
15.0%
23.1%
16.5%
16.0%
Adjusted R
F-statistic
1.5 ***
1.5 ***
1.6 ***
1.6 ***
1.4 **
Obs.
552
828
368
552
368
Sector (Panel A) / Company (Panel B) and Year Fixed Effects Included in all equations
0.037
0.0094
0.009
18.2%
1.6 ***
552
42
0.049
0.0776 ***
0.010
28.6%
1.8 ***
552
0.032
0.0647 ***
0.009
24.6%
1.9 ***
828
0.062
19.4%
1.5 **
(45)
n= -1 to 0
-0.2149 *
0.121
0.0769
0.182
-0.0020
0.012
11.6%
1.9 ***
(46)
n= 1 to 3
-0.3611 **
0.121
0.5312 ***
0.182
0.1219 ***
0.041
0.041
0.1069 **
-0.0665 *
0.054
0.0786 ***
0.011
34.2%
2.0 ***
368
0.054
0.0615 ***
0.011
28.2%
2.1 ***
552
0.047
28.3%
1.9 ***
(47)
n= -1 to 0
-0.0862
0.181
-0.0976
0.237
0.1046 **
0.006
0.0251 **
0.010
9.8%
1.7 ***
(48)
n= 1 to 3
-0.1823
0.162
0.3624 **
0.181
0.0419
0.048
0.032
0.0088
0.0106
0.058
0.0656 ***
0.013
26.3%
1.7 ***
368
0.039
0.0501 ***
0.009
26.5%
2.0 ***
552
Table 9 (Continued)
Last alternative Investment intensity measure: CAPEX deflated with Sales.
CAPEX / Sales
All
PANEL A
Period
ROA Lag
(49)
n= -1
-0.3036
0.236
Sales g
0.4191 ***
PE dummy
0.238
0.6595 ***
0.123
0.0912
0.064
0.137
0.072
0.158
0.067
0.0132
0.0046
-0.0301
-0.0025
17.3%
1.7 ***
(55)
n= -1 to 0
0.2349
0.568
0.017
0.3216 ***
0.110
16.3%
2.9 ***
(56)
n= 1 to 3
-0.4559
0.033
0.019
0.031
0.3805 **
0.2940 **
0.4106 *
0.188
14.7%
1.4 *
(57)
n= -1 to 0
-0.2866
0.282
-0.2065
Sales g
-0.0200
0.120
0.103
0.168
x PE
0.1156
-0.0419
0.2711
0.621
0.168
0.1053 **
0.047
0.8214 **
0.3038
0.359
0.1854 *
0.145
20.3%
2.7 ***
(58)
n= 1 to 3
-0.7236 *
0.359
x PE
Constant
0.137
0.2769 ***
0.0000
0.5005 ***
2
0.364
0.1286
0.125
0.143
Adjusted R
F-statistic
PANEL B
Period
ROA Lag
0.112
0.2100 ***
PE + PTA
(53)
(54)
n= -1
n= 1 to 3
-0.4232 *** -0.3128 **
-0.0093
0.028
Constant
PE + PROA
(51)
(52)
n= -1
n= 1 to 3
0.6517 *
-0.4733 ***
(50)
n= 1 to 3
-0.4498 ***
0.378
1.0370 **
0.437
-0.1871
0.126
0.428
0.4035 ***
0.122
-0.2521 *
0.201
0.1437 ***
0.1431 ***
0.023
0.142
0.1387 ***
0.029
0.027
2
42.1%
32.8%
38.5%
38.6%
Adjusted R
F-statistic
2.4 ***
2.4 ***
2.2 ***
2.8 ***
Obs.
552
828
368
552
Sector (Panel A) / Company (Panel B) and Year Fixed Effects Included in all equations
0.239
25.5%
1.8 ***
(59)
n= -1 to 0
1.1649
0.018
0.4329 ***
0.132
11.8%
1.9 ***
(60)
n= 1 to 3
-0.0195
0.807
0.425
-1.1168
0.3530
0.831
0.474
0.1642
-0.1597
0.130
-0.0628
0.169
0.0508
0.049
40.8%
2.3 ***
368
0.124
0.2767 *
0.142
0.1045 ***
0.026
36%
2.6 ***
552
This impact can be seen in two ways. The first impact is seen in sensitivity to
investment opportunities. When measured against peers matched on ROA, for all of the
CAPEX intensity measures, after the PE entry, the firms become less sensitive to
investment opportunities, as it can be seen by the negative sign of the coefficient which
would add (subtract in this case) to the generic coefficient. Furthermore, in some cases
the PE was more sensible than its peers before entry.
For example, the CAPEX/Lag TA of a PE firm before PE entry would typically have a
15.27 pp (0.1527) higher sensitivity to investment opportunities than its peers with
similar ROA (regression 15), whilst after the entry, the value was 8.58 pp (-0.0858)
lower (regression 16).
Nevertheless, the results are not the same if we match firms based on size (TA) rather
than ROA. This brings us to the question on what is the better matching alternative. We
tend to consider ROA as the better matching alternative.
43
First, in the literature, a statistical significant relationship between ROA and investment
intensity is systematically reported, and hence its inclusion in investment regressions
even for authors agnostic to its interpretation (Asker et al., 2015). Size does not seem to
have a similar impact in explaining investment policy.
Second, we could match on ROA and TA. However as discussed under the
Methodology section, besides the overmatching argument, our sample of private firms
does not include sufficient companies to hold the threefold matching criteria. We’d end
up with very different companies under one of the dimensions or would have to relax
the level of industry matching.
Furthermore, the first matching would impact the final result, and the second matching
would only refine the results, which would eventually be different all the same, meaning
that matching on TA and then ROA would probably result in different outcomes than
matching on ROA and then TA.
Third, despite the fact that both matching criteria result in non statistical differences in
investment intensities with PE firms in selection year (n-1), the distribution in ROA
firms seems less subjected to outliers and more centred (Figure 3). Additionally, the
average difference in matching (under each criterion) seems smaller in ROA, meaning
that we find on average more firms with a very similar ROA than firms with a very
similar TA (under each sector).
Finally, the sensitivity to investment opportunities (not the PE effect) is almost always
statistically more significant in regressions made in a sample of PE and PROA firms
than in samples of PE and PTA firms.
Nonetheless, this is clearly a subject that would require additional research and probably
deserves a separate research topic.
The second evidence of a negative impact of PE in investment is given by the sensitivity
to ROA.
In this case, the impact is more or less consistent across the metrics and matching
criteria, with the difference being that in PTA sample, the value is only significant when
44
CAPEX or expansionary CAPEX is measured against end-of-year assets (regressions 42
and 48).
The rather consensual result is that sensitivity to ROA, or in other words, the ICF
sensitivity, increases and is positive and statistically significant for PE firms after the
entry.
Despite divergence on the interpretation towards the ICF, we recall our literature
review, namely Bertoni et al. (2013), who stated that the Kaplan and Zingales (1997)
critique is about the monotonicity not about its sign, meaning that the ICF sensitivity
should not be a measure of the degree of financial constraints but a sign of its existence.
Hence, non-financially constrained firms should have an ICF not statistically different
from zero but financially constrained firms, would have positive and significant ICF
sensitivities.
This is precisely our case. The only statistically significant positive sensitivities to
ROA, in our sample/regressions, are the ones regarding PE firms.
In fact, we saw a significant increase in leverage towards PROA firms. By year 3, the
average difference in PE firms Debt/EBITDA to the Sector median was 5.8x whilst the
same difference in the sample of firms matched on ROA was 2.8x.
All in all, the results seem to indicate the existence of financial constraints in PE backed
firms, which appear only after entry, that end up underinvesting its comparable peers.
This seems consistent with Ughetto (2014) findings.
45
5. The Public vs. Private discussion
5.1 The discussion and relevance for our analysis
Is there a difference in investment policies between public and private firms, or, in other
words, does the listing status impact a firm investment intensity?
As we saw in our literature review, the theoretical answer to this question is not
consensual. Whilst according to the free cash flow hypothesis (Jensen, 1986), on the
back of agency related issues one would expect public firms to overinvest in relation to
similar private firms; Stein (1988), however, offers a different view, arguing that the
public firm focus on short term results, which often leads to myopically cuts in
investment, sacrificing long term objectives due to short term pressures.
This, as we saw, is relevant for our analysis as the implicit or explicit explanation for a
majority of the observed reduction in investment after a going private, or otherwise PE
intervention in general, is that it simply results from a correction in overinvestment.
As we mentioned previously, very recently Asker et al. (2015) showed empirical
evidence in support of Stein (1988) arguments, for the US.
5.2 Empirical Findings
We envisaged presenting a high level comparison for Europe and thus we did a similar,
albeit much less extensive, approach to Asker et al. (2015)1.
We present in Table 10 the results for the unconditional investment intensities, as
measured by CAPEX / Lagged total assets, for the 9 years from 2005 to 2013.
1
Please bear in mind that this approach is mainly related to our PE deals sample and, thus, the publicprivate sample encompasses only the sectors analysed in our main study and not the whole economy.
Nevertheless, we seem to have a representative sample of the European economy, given we have 76
different Nace 4-digit sectors, encompassing all the 1 digit codes and the fact that our sample of 29,435
companies, represents €3.3tn in assets and €2.4tn sales (2013), which compares with the Euro-28 €13.9tn
GDP in 2014. As discussed in methodology, we envisaged a stable sample, meaning the same set of firms
throughout the analysis period. Hence we discarded all the companies lacking any of our main financials
during any year of the analysed period. Between lacking some or all the financials, from a starting almost
400 thousand companies, we ended up with a sample of 29 thousand companies, before matching, as
represented in Appendix III.
46
Table 10 - Unconditional Investment Intensities
This table reports unconditional investment intensities for public and private firms per year, in
four samples: all, unmatched, and three matched on Nace 4-digit code and total assets, ROA and
total assets plus ROA, respectively. Significance tests for the difference between public and
private samples given Anova F-Test and the two-tailed Wilcoxon rank sum for means and
medians, respectively. We use ***, **, and * to denote significance at the 1%, 5%, and 10% level
(two-sided), respectively. Values are winsorised between 0.005 and 0.995 percentiles to reduce
the impact of outliers. We use 0.5% in each tail rather than 5% as previously as the outliers carry
less weight given the size of the samples and to compare with Asker et al. (2015). For the cases
where there is a statistically significant difference, we highlight in bold the higher value.
CAPEX / Lag TA Nº Firms
2005
2006
2007
2008
2009
2010
2011
2012
2013
MEANS
All Sample
Private
Public
All
Difference
17,803
11,632
29,435
0.0829
0.0917
0.0863
***
0.0800
0.0854
0.0821
***
0.0693
0.0840
0.0751
***
0.1225
0.1272
0.1244
0.0521
0.0564
0.0564
0.0641
0.0583
0.0618
***
0.0515
0.0534
0.0522
*
0.0461
0.0483
0.0470
**
0.0363
0.0423
0.0387
***
Matched on TA
Private
Public
All
Difference
9,306
9,306
18,612
0.1037
0.0917
0.0977
***
0.0892
0.0899
0.0896
0.0705
0.0873
0.0789
***
0.1345
0.1409
0.1377
0.0623
0.0588
0.0605
*
0.0690
0.0619
0.0654
***
0.0551
0.0571
0.0561
0.0502
0.0505
0.0503
0.0391
0.0443
0.0417
***
Matched on ROA
Private
Public
All
Difference
9,596
9,596
19,192
0.0989
0.0905
0.0947
***
0.0955
0.0837
0.0896
***
0.0829
0.0831
0.0830
0.1757
0.1345
0.1551
***
0.0687
0.0562
0.0625
***
0.0719
0.0590
0.0654
***
0.0608
0.0536
0.0572
***
0.0536
0.0492
0.0514
***
0.0423
0.0433
0.0428
0.102
0.0867
0.0942
***
0.0903
0.0865
0.0884
0.0704
0.0850
0.0777
***
0.1313
0.1415
0.1364
*
0.0626
0.0575
0.0601
**
0.0700
0.0615
0.0658
***
0.0561
0.0566
0.0564
0.0498
0.0501
0.0500
0.0395
0.0432
0.0413
**
Matched on TA & ROA
Private
8,393
Public
8,393
All
16,786
Difference
MEDIANS
All Sample
Private
Public
All
Difference
17,803
11,632
29,435
0.0343
0.0371
0.0355
***
0.0363
0.0341
0.0355
***
0.0292
0.0346
0.0313
***
0.0317
0.0391
0.0349
***
0.0244
0.0224
0.0235
***
0.0281
0.0252
0.0270
***
0.0221
0.0235
0.0226
0.0199
0.0213
0.0206
0.0136
0.0186
0.0156
***
Matched on TA
Private
Public
All
Difference
9,306
9,306
18,612
0.0364
0.0368
0.0366
0.0376
0.0345
0.0361
***
0.0281
0.0348
0.0313
***
0.0283
0.0395
0.0342
***
0.0257
0.0224
0.0240
***
0.0300
0.0254
0.0276
***
0.0232
0.0235
0.0235
0.0222
0.0216
0.0219
0.0148
0.0187
0.0169
***
Matched on ROA
Private
Public
All
Difference
9,596
9,596
19,192
0.0358
0.0380
0.0368
**
0.0387
0.0350
0.0369
***
0.0313
0.0354
0.0334
***
0.0327
0.0405
0.0368
***
0.0263
0.0228
0.0245
***
0.0304
0.0255
0.0278
***
0.0237
0.0238
0.0238
0.0221
0.0214
0.0217
0.0155
0.0189
0.0171
***
0.0368
0.0366
0.0367
0.0387
0.0347
0.0367
***
0.0286
0.0350
0.0316
***
0.0290
0.0401
0.0348
***
0.0265
0.0223
0.0243
***
0.0309
0.0255
0.0279
***
0.0242
0.0237
0.0239
0.0231
0.0216
0.0223
0.0154
0.0186
0.0170
***
Matched on TA & ROA
Private
8,393
Public
8,393
All
16,786
Difference
47
The whole sample, unmatched, shows that public firms, on average, invest more than
private firms. This happens in 6 out of the 9 years of the sample. In 2008/9 there is no
(significant) difference and in 2010 private firms invest more.
Samples matched on TA and TA & ROA2 show broadly inconclusive results, as there as
many years where public firms invest more as the other way around.
Again, like in our PE sample, here matching on ROA also shows a different story, in
this case, that in the majority of years private firms invest more than public firms (7 out
of 9, always significant at 1%) and there is no year where the public firms invest
significantly more than their private counterparts (in mean terms).
Curiously, all the samples show a very similar story in median terms: before the crisis,
in 2006 private firms invested more, then in 2007/08 public firms invest more than
private firms, and then in the two following years the situation reverts again (at least
temporarily).
This evolution is basically conditioned by the quick fall in investment intensity in
private firms as public firms seem rather unaffected by the crisis, at least in its early
years (public firm’s investment also reduces but only 2 years after private peers, in
2009). This is consistent with the view public firms have broader access to capital
(equity and debt) markets, whilst private firms are much more dependent on the banking
sector for external financing, which, as it is common knowledge, was severely halted in
the wake of the financial crisis.
As shown in Table 11, Panel A (estimation of equation 3.1), holding investment
opportunities (sales growth as proxy) and profitability constant for the whole period, the
sample matched on ROA shows that public firms invest less than their private
counterparts, which is consistent Asker et al. (2015) conclusions. Typically, a public
firm investment intensity would be 0.6 pp less (-0.0061) than the private counterpart,
with the same profitability and investment opportunities.
2
Unlike with our PE sample, here we can perform a matching on TA and ROA (much larger sample).
However, given the fact that the matching first occurs on TA and then on ROA, the results are
undoubtedly conditioned by the first step and the second step is nothing more than refining the first one.
Hence, the first step can lead to the exclusion of peers which could under another criterion be classified as
better comparable.
48
Table 11 - Conditional Investment Intensities
This table reports the estimation of Equation (3.1) in Panel A and Equation (3.2) in Panel B. Whilst Equation (3.1) isolates the public listing status, Equation (3.2)
and allows the analysis of within-firm variation to differences in the sensitivity of investment intensity to investment opportunities, with sales growth as proxy, and
profitability, between public and private. We estimate the regression of investment intensity as measured by CAPEX / Lagged total assets in three matched samples
and for 4 estimation periods, the whole period, 2005/06 as “pre-crisis” period, 2007/10 as “crisis period” and 2011/13 as the “port crisis". In Panel A the regressions
include Sector (Nace 4-digit code) and year fixed effects as the specification does not allow for company fixed effects. In Panel B company and year fixed effects
are included. Fixed Effects test given by Redundant Fixed Effects – Likelihood Ratio. Heteroskedasticity-consistent standard errors (White diagonal) are shown in
italics under the coefficient estimates. We use ***, **, and * to denote significance at the 1%, 5%, and 10% level (two-sided), respectively. Values are winsorised
between 0.005 and 0.995 percentiles (as in Asker et al. (2015)).
Period
Obs.
Companies
PANEL A
Public
Matched on Sector & ROA
2005 - 2013 2005-2006 2007-2010 2011-2013
172,695
19,192
38,372
19,191
76,756
19,192
57,567
19,192
-0.0061 *** -0.0058 *** -0.0089 *** -0.0024 **
0.001
0.002
0.001
0.001
ROA Lag
0.1932 *** 0.2402 *** 0.1821 *** 0.1792 ***
Sales g
0.0711 *** 0.0773 *** 0.0767 *** 0.0554 ***
Constant
0.0488 *** 0.0429 *** 0.0597 *** 0.0377 ***
0.006
0.003
0.003
Sector and Year FE:
Adjusted R2
F-statistic
PANEL B
ROA Lag
Yes ***
7.9%
174.0 ***
0.051
0.072
0.008
0.0114 ** -0.0025
0.005
Contant
0.004
0.005
Yes ***
7.6%
78.8 ***
0.009
0.004
0.005
Yes ***
6.6%
51.8 ***
0.024
0.025
-0.1056 *** -0.0122
0.033
0.043
0.0513 *** 0.0287 *** 0.0362 *** 0.0325 ***
0.003
x Public
0.006
Yes ***
8.3%
44.8 ***
-0.0849 *** -0.0540
0.019
Sales g
0.005
0.009
0.2839 *** 0.4324 *** 0.3908 *** 0.1921 ***
0.013
x Public
0.013
0.0443
0.001
0.012
0.006
0.0163 *
0.009
0.006
0.0022
0.009
0.0309 *** 0.0446 *** 0.0336 ***
0.005
Company and Year FE: Yes *** Yes ***
Adjusted R2
19.2%
10.9%
F-statistic
2.10 ***
1.48 ***
0.002
0.002
Yes *** Yes ***
9.8%
15.1%
1.44 ***
1.53 ***
Matched on Sector & Total Assets
2005 - 2013 2005-2006 2007-2010 2011-2013
167,439
18,612
0.0001
37,197
18,612
74,423
18,611
-0.0051 *** 0.0017
0.001
0.002
0.001
55,819
18,611
0.013
0.008
-0.0005
0.001
0.001
0.009
0.005
0.004
0.004
0.003
0.006
0.006
0.004
Yes ***
7.0%
70.0 ***
0.0767 ***
0.013
33,565
16,786
-0.0061 ***
0.002
0.2196 ***
0.013
0.0750 ***
0.005
0.0480 ***
0.017
67,136
16,786
50,348
16,786
0.0015
0.0007
0.001
0.001
0.1695 *** 0.1737 ***
0.009
0.009
0.0748 *** 0.0513 ***
0.004
0.004
0.0894 *** 0.0773 ***
0.023
0.019
Yes ***
6.2%
47.2 ***
Yes ***
7.8%
150.9 ***
Yes ***
8.1%
39.2 ***
Yes ***
7.3%
66.8 ***
0.1957 *** 0.3242 *** 0.2813 *** 0.1615 ***
0.2411 ***
0.4576 ***
0.3230 *** 0.1554 ***
0.013
Yes ***
7.7%
40.3 ***
0.0686 ***
0.003
0.0521 *** 0.0556 *** 0.0643 *** 0.0334 ***
Yes ***
7.4%
157.6 ***
0.1809 ***
0.006
0.0685 *** 0.0792 *** 0.0719 *** 0.0490 ***
0.003
151,049
16,786
0.0014
0.1507 *** 0.1476 *** 0.1498 *** 0.1632 ***
0.006
Matched on Sector & Total Assets & ROA
2005 - 2013 2005-2006 2007-2010 2011-2013
0.047
0.022
0.025
-0.0501 *** -0.1007
-0.0236
-0.0069
0.031
0.039
0.018
0.065
0.003
0.005
0.009
0.005
0.0080
0.0080
0.012
0.008
0.005
0.0170 **
0.008
0.00
0.00
49
0.00
-0.0652 ***
-0.175 **
0.0514 ***
0.025
0.0056
0.033
0.041
0.0429 *** 0.0271 ***
0.003
0.009
0.005
0.005
-0.002
0.0044
0.0090
0.008
0.009
0.013
0.0455 ***
0.0356 ***
Yes ***
11.3%
2.15 ***
Yes ***
20.0%
1.50 ***
0.002
Yes ***2 Yes ***
9.5%
14.5%
1.42 ***
1.51 ***
0.070
0.0271 ***
0.023
-0.0379
0.0072
0.005
0.0515 *** 0.0533 *** 0.0503 *** 0.0357 ***
Yes ***1 Yes ***4
10.9%
19.8%
2.09 ***
1.49 ***
0.050
0.018
0.0504 *** 0.0250 *** 0.0414 *** 0.0238 ***
0.0121 **
0.013
Yes ***
6.6%
45.9 ***
0.001
0.0440 *** 0.0348 ***
0.005
Yes ***
9.9%
1.44 **
*
0.002
Yes ***
14.6%
1.51 ***
This holds out through the subsamples we’ve created for the “pre-crisis” (2005/6),
“crisis” (2007/10) and “post-crisis” (2011/13) periods. Curiously, the difference in
public firms towards private peers increases in the “crisis” period (-0.89 pp).
However, matching both on TA and TA&ROA shows no statistically significant
difference for the public vs. private companies in the whole period, after controlling for
investment opportunities and profitability.
Nonetheless, looking to the subsamples, we can see that in the pre-crisis period, there
was a difference, which was consistent with the sample matched on ROA, i.e., the
public firms invest less than their private peers and the value is quite similar across the
three matching criteria.
With the crisis, and after it, the difference disappears. This could signal both that the
impact of the crisis was still present in our otherwise classified as “post-crisis” period
(in fact if we consider it just as one period the results are the same), or that there was a
change in the investment structure across firms.
Albeit this would require further investigation, out of the scope in this study, this can
relate to the aforementioned dependency of European private firms to banking
financing, unlike their US counterparts, much more relying on capital markets.
This also seems consistent with the results from Equation (3.2) as reported in Table 11,
Panel B, which as we referred, splits the “public effect” of equation (3.1), as seen in
Panel A, between the profitability/cash-flow and investment opportunities explanatory
variables.
Although with some differences in sub periods, for the whole period analysed, all three
matching criteria output the a consistent idea, which is that the impact of the public
listing status is basically due to the lower ICF sensitivity as, unlike Asker et al. (2015),
if any, the difference in sensitivity to investment opportunities is higher in public firms.
For instance, in a sample of matched peers on sector and TA, the public firm sensitivity
to
investment
opportunities,
for the whole period (2005/13), was 0.0625
(0.0504+0.0121) whilst the private firm was 0.0504. This compares with the 0.028 and
0.118, respectively estimated by Asker et al. (2015) for the US in the 2002/11 period.
50
Looking at sub periods, it seems that the same outcome in the ROA and in the TA
matching samples have different origins, as the difference in investment opportunities
seems to arise from the “crisis” period in the former and in the “post-crisis” in the latter.
The sample matched by TA&ROA produces no difference in investment opportunities
across public or private firms.
Although subjected to discussion, the lower ICF sensitivity in public firms seems
consistent with the fact that, at least in the “pre-crisis” period, public firms did invest
less than private firms and that with the “crisis”, private firms reduce more their
investment.
If private firms are more dependent on the banking sector for financing and have less
access to capital markets, then it’s plausible that they can be more dependent on cashflow to finance its investments. And in fact, with the profitability/cash-flow plunge
during the crisis, private firms’ investment is severely affected whilst public firms’ is
not, or, at least, is highly lagged (2 years). We do not show these values, but to
illustrate, the mean private (public) ROA Lag was 13.3% (12.2%) in 2007 and falls to
10% (9%) in 2009.
European firms have historically been more dependent on banking, which can also
explain some of the differences to the US. In fact, a special report from FitchRatings
(2013) shows an increase of the bond weight in European corporate funding from a 17%
in 2005 to 52% in 2013 (1H). One of the reasons presented is the banking deleveraging
trend, on the back of regulatory pressures (e.g. Basel III). Nonetheless, despite the
evolution, European firms are still long way until the typical American capital structure
where bonds represent [70-80]% of the whole debt.
All-in-all, our results show higher sensitivity to ROA than to investment opportunities,
while for US, Asker et al. (2015) found opposite results.
51
6. Conclusions
Only recently the overinvestment correction (Jensen, 1986), as an explanation for the
evidence that CAPEX falls after the PE entry, started to be questioned, with Sousa and
Jenkinson (2013), Bharath et al. (2014) and Ughetto (2014) concluding that evidence is,
at least, not supportive of that hypothesis, and Asker et al. (2015) showing that in US
public firms invest less and are less sensitive to changes in investment opportunities
than private firms.
Our research goal was twofold: first assess the impact of PE entry in European
companies, in recent years, and second, try to compare the investment policies of public
and private firms, in order to briefly compare to Asker et al. (2015) results for US with
European data, and verify empirically the question of public firm overinvestment
(Jensen, 1986) or underinvestment (Stein, 1988) thesis. Both goals are related, as the
common explanation for the investment intensity reduction after the buyout is exactly
the correction of an overinvestment.
Using Zephyr and Amadeus databases, we have collected a sample of 92 PE entry deals
in Europe-28, between 2006 and 2010, and a sample of c. 29 thousand European
companies (c. 11 thousand public and 18 thousand private) and tried to answer to the
questions: How do investments evolve after the PE entry a what can we conclude from
that evolution after controlling matched peers and for the variables that according to the
empirical investment literature explain the investment intensity? Do European public
firms overinvest their private comparables?
Using the Kaplan (1989) approach we’ve compared the pre-entry (n-1) with the post
entry (n+1 to n+3) investment levels and found some evidence that PE firms invest
statistically significant less than their peers, in a period marked by the crisis and where
there was a generic trend (sector medians) to reduce investment intensity.
However, this trend is not consistent across different matching criteria and appears only
when ROA (after sector) is our matching criteria.
Controlling for the investment intensity standard explainable variables, found in the
empirical investment literature, and following Asker et al. (2015) methodology, we
52
found some evidence that only PE firms exhibit positive significant ICF sensitivity. This
results, consistent with Ughetto (2014), seem to indicate the existence of financial
constraints in PE backed firms.
The literature critiques on the ICF sensitivity interpretation relate to the fact that it
should not be considered a measure of the degree of financial constraints but a sign of
its existence: financially constrained firms would have positive and significant ICF
sensitivities (Bertoni et al., 2013).
We also found some evidence that after the PE entry, firms become less sensitive to
investment opportunities, with sales growth as proxy, which unlike the ICF, is
dependent on the matching criteria, in the case, only occurring when peers are matched
by ROA. Despite acknowledging that the question can be debatable, and would require
further investigation, we present some arguments that support ROA rather than total
assets as matching criteria.
Finally we’ve compared public to private firms in Europe during the last 10 years.
We found some evidence that private firms invested more than their public counterparts
before and, under some criteria/matching did it again after the crisis, even if temporarily
in some cases. The convergence, during the early crisis years, was caused by significant
fall in investment in private firms and not by an increase in public firms (which lag the
private counterparts drop by 2 years).
At the same time, we found lower ICF sensitivity in public firms, which seems
consistent with the fact that, with the initial stage (2007/08) of the “crisis”, private firms
reduce their investment whilst public firms did not. If private firms are more dependent
on banking financing and have less access to capital markets, then it’s plausible that
they can be more dependent on cash-flow to finance their investments.
Although with some differences in sub periods, for the whole period analysed, all three
matching criteria output the same consistent idea, which is that the impact of the public
listing status – it reduces, at least before the crisis, the investment intensity - is basically
due to the lower ICF sensitivity as, unlike Asker et al. (2015), if any, the difference in
sensitivity to investment opportunities is higher in public firms.
53
In a nutshell, we find some evidence, even though limited, that PE impact negatively
firms investment policies due to a mix of increased financial constraints and probably to
lower sensitivity to investment opportunities. In any case, we found stronger evidence
that the overinvestment correction is hardly a valid explanation, as public firms, at least
before the crisis, invested less than their private counterparts.
This research has some limitations. The concentration of deals in the “crisis” period
seems to considerably “taint” our sample, the ultimate example being the fact that
unlike any other research we found no evidence of profitability increase after the entry
(by n+3). In fact, the period is marked by a significant generalized reduction in
investment and profitability, which can to some extent “mask” the PE firm behaviour.
Furthermore, our sample of public and private firms, albeit significantly larger than in
Asker et al. (2015), does not encompass all the sectors nor European countries.
This research leaves several topics for future research, namely: the development of our
PE sample, in order to be possible to isolate the “crisis” effect; the analysis of the exit
(we were not able to pursue due to the lack of deals with data); a deeper insight on what
matching criteria produces better and more consistent results, under more “normalized”
economic circumstances; the enlarging of the public/private sample to the whole set of
sector codes, the countries influence, as well as a deeper investigation on the differences
and its motives to the US.
54
References
Achleitner, A.-K., and Figge, C. (2014), "Private Equity Lemons?Evidence on Value
Creation in Secondary Buyouts", European Financial Management, Vol. 20, Nº 2, pp.
406-433.
Asker, J., Farre-Mensa, J., and Ljungqvist, A. (2015), "Corporate Investment and Stock
Market Listing: A Puzzle?", Review of Financial Studies, Vol. 28, Nº 2, pp. 342-390.
Bertoni, F., Ferrer, M., and Martí, J. (2013), "The different roles played by venture
capital and private equity investors on the investment activity of their portfolio firms",
Small Business Economics, Vol. 40, Nº 3, pp. 607-633.
Bharath, S., Dittmar, A., and Sivadasan, J. (2014), "Do Going-Private Transactions
Affect Plant Efficiency and Investment?", Review of Financial Studies.
Bharath, S. T., and Dittmar, A. K. (2010), "Why Do Firms Use Private Equity to Opt
Out of Public Markets?", Review of Financial Studies.
Boucly, Q., Sraer, D., and Thesmar, D. (2011), "Growth LBOs", Journal of Financial
Economics, Vol. 102, Nº 2, pp. 432-453.
Bruton, G. D., Keels, J. K., and Scifres, E. L. (2002), "Corporate restructuring and
performance: An agency perspective on the complete buyout cycle", Journal of
Business Research, Vol. 55, Nº 9, pp. 709-724.
Campello, M., and Graham, J. R. (2013), "Do stock prices influence corporate
decisions? Evidence from the technology bubble", Journal of Financial Economics,
Vol. 107, Nº 1, pp. 89-110.
Chauvin, K. W., and Hirschey, M. (1993), "Advertising, R&D Expenditures and the
Market Value of the Firm", Financial Management, Vol. 22, Nº 4, pp. 128-140.
Chung, J.-W., (2011), "Leveraged buyouts of private companies", working paper.
Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1904342.
Economics, F. (2013). Exploring the impact of private equity on economic growth in
Europe. In EVCA (Ed.), EVCA Research.
Engel, D., and Stiebale, J. (2014), "Private equity, investment and financial constraints:
firm-level evidence for France and the United Kingdom", Small Business Economics,
Vol. 43, Nº 1, pp. 197-212.
55
Fazzari, S. M., Hubbard, R. G., and Petersen, B. C. (2000), "Investment-Cash Flow
Sensitivities are Useful: A Comment on Kaplan and Zingales", The Quarterly Journal
of Economics, Vol. 115, Nº 2, pp. 695-705.
Fazzari, S. M., Hubbard, R. G., Petersen, B. C., Blinder, A. S., and Poterba, J. M.
(1988), "Financing Constraints and Corporate Investment", Brookings Papers on
Economic Activity, Vol. 1988, Nº 1, pp. 141-206.
FitchRatings (2013). European Corporate Funding Disintermediation Special Report:
Fitch Ratings.
Gilligan, J., and Wright, M. (2012). Private Equity Demystified: 2012 Update. In I. C.
F. Faculty (Ed.), Private Equity Demystified: ICAEW.
Graham, J. R., Harvey, C. R., and Rajgopal, S. (2005), "The economic implications of
corporate financial reporting", Journal of Accounting and Economics, Vol. 40, Nº 1–3,
pp. 3-73.
Guo, S., Hotchkiss, E. S., and Song, W. (2011), "Do Buyouts (Still) Create Value?",
The Journal of Finance, Vol. 66, Nº 2, pp. 479-517.
Hadlock, C. J. (1998), "Ownership, Liquidity, and Investment", The RAND Journal of
Economics, Vol. 29, Nº 3, pp. 487-508.
Hall, B. H., Berndt, E., and Levin, R. C. (1990), "The Impact of Corporate
Restructuring on Industrial Research and Development", Brookings Papers on
Economic Activity. Microeconomics, pp. 85-135.
Holthausen, R. W., and Larcker, D. F. (1996), "The financial performance of reverse
leveraged buyouts", Journal of Financial Economics, Vol. 42, Nº 3, pp. 293-332.
Jensen, M. C. (1986), "Agency Costs of Free Cash Flow, Corporate Finance, and
Takeovers", The American Economic Review, Vol. 76, Nº 2, pp. 323-329.
Jensen, M. C. (1989), "Eclipse of the Public Corporation", Harvard Business Review,
Vol. 67, Nº 5, pp. 61-74.
Kaplan, S. (1989), "The effects of management buyouts on operating performance and
value", Journal of Financial Economics, Vol. 24, Nº 2, pp. 217-254.
Kaplan, S. N., and Zingales, L. (1997), "Do Investment-Cash Flow Sensitivities Provide
Useful Measures of Financing Constraints?", The Quarterly Journal of Economics, Vol.
112, Nº 1, pp. 169-215.
56
Kaplan, S. N., and Zingales, L. (2000), "Investment-Cash Flow Sensitivities Are Not
Valid Measures of Financing Constraints", The Quarterly Journal of Economics, Vol.
115, Nº 2, pp. 707-712.
Kerstein, J., and Kim, S. (1995), "The Incremental Information Content of Capital
Expenditures", The Accounting Review, Vol. 70, Nº 3, pp. 513-526.
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., and Vishny, R. (2002), "Investor
Protection and Corporate Valuation", The Journal of Finance, Vol. 57, Nº 3, pp. 11471170.
Lerner, J., Sorensen, M., and Stromberg, P. (2011), "Private Equity and Long-Run
Investment: The Case of Innovation", The Journal of Finance, Vol. 66, Nº 2, pp. 445477.
Lichtenberg, F. R., and Siegel, D. (1990), "The effects of leveraged buyouts on
productivity and related aspects of firm behavior", Journal of Financial Economics,
Vol. 27, Nº 1, pp. 165-194.
Long, W. F., and Ravenscraft, D. J. (1993), "LBOs, debt and R&D intensity", Strategic
Management Journal, Vol. 14, Nº S1, pp. 119-135.
McConnell, J. J., and Muscarella, C. J. (1985), "Corporate capital expenditure decisions
and the market value of the firm", Journal of Financial Economics, Vol. 14, Nº 3, pp.
399-422.
Morck, R., Shleifer, A., and Vishny, R. W. (1988), "Management ownership and market
valuation: An empirical analysis", Journal of Financial Economics, Vol. 20, Nº 0, pp.
293-315.
Moyen, N. (2004), "Investment–Cash Flow Sensitivities: Constrained versus
Unconstrained Firms", The Journal of Finance, Vol. 59, Nº 5, pp. 2061-2092.
Muscarella, C. J., and Vetsuypens, M. R. (1990), "Efficiency and Organizational
Structure: A Study of Reverse LBOs", The Journal of Finance, Vol. 45, Nº 5, pp. 13891413.
Myers, S. C., and Majluf, N. S. (1984), "Corporate financing and investment decisions
when firms have information that investors do not have", Journal of Financial
Economics, Vol. 13, Nº 2, pp. 187-221.
Poterba, J. M., and Summers, L. H. (1995), "A CEO survey of US companies' time
horizons and hurdle rates", Sloan Management Review, Vol. 37, pp. 43-43.
57
Rappaport, A. (1990), "The staying power of the public corporation", Harvard Business
Review, Vol. 68, Nº 1, pp. 96-104.
Shapiro, R. J., and Pham, N. D., (2009), "The impact of private equity acquisitions and
operations on capital spending, sales, productivity, and employment", Paper presented
at the The Private Equity Council,
Smith, A. J. (1990), "Corporate ownership structure and performance: The case of
management buyouts", Journal of Financial Economics, Vol. 27, Nº 1, pp. 143-164.
Sousa, M., and Jenkinson, T., (2013), "Keep Taking the Private Equity Medicine? How
Operating Performance Differs between Secondary Deals and Companies that Go to
Public Markets", working paper. Retrieved from http://ssrn.com/abstract=2019595.
Stein, J. C. (1988), "Takeover Threats and Managerial Myopia", Journal of Political
Economy, Vol. 96, Nº 1, pp. 61-80.
Strömberg, P., (2009), "The economic and social impact of private equity in Europe:
summary
of
research
findings",
working
paper.
Retrieved
from
http://ssrn.com/abstract=1429322.
Ughetto, E. (2010), "Assessing the contribution to innovation of private equity
investors: A study on European buyouts", Research Policy, Vol. 39, Nº 1, pp. 126-140.
Ughetto, E. (2014), "Investments, Financing Constraints and Buyouts: the Effect of
Private Equity Investors on the Sensitivity of Investments to Cash Flow", The
Manchester School, pp. n/a-n/a.
Weisbach, M. S. (1988), "Outside directors and CEO turnover", Journal of Financial
Economics, Vol. 20, Nº 0, pp. 431-460.
Wright, M., Gilligan, J., and Amess, K. (2009), "The economic impact of private equity:
what we know and what we would like to know", Venture Capital, Vol. 11, Nº 1, pp. 121.
58
Appendix I
Table 12 - Main Studies addressing CAPEX impact on PE backed firms
Author/Year
of study
Kaplan
(1989)
Smith (1990)
Muscarella
and
Vetsuypens
(1990)
Holthausen
and Larcker
(1996)
Boucly et al.
(2011)
(Chung,
2011)
Sousa and
Jenkinson
(2013)
Bertoni et al.
(2013)
Engel and
Stiebale
(2014)
Bharath et al.
(2014)
Ughetto
(2014)
Sample
Findings
Underinvestment vs Overinvestment
Hypothesis
76 US MBOs
CAPEX falls in all 3 years after buyout Suggests that indirect evidence given by
between 1980-86 although not statistically significant.
the fact that market adjusted returns to
Industry adjusted reductions are larger
post-buyout investors is large and
and significant.
significant points to the overinvestment
hypothesis.
58 US MBOs
CAPEX significantly falls after buyout.
Not addressed
between 1977-86 CAPEX to Sales also decreases but is
not the major cause for the increase in
RoA
72 RLBOs from
RLBO has CAPEX/Sales lower than
Not addressed
1983-87 of LBOs industry and experienced a decrease in
occurred 1976-86
the relative level of CAPEX under
private ownership. Decline is most
relevant for subsample not engaging in
acquisition/divestiture activity.
90 RLBOs from Pre IPO LBOs have lower CAPEX than Authors refer to the fact that the increase
1983-88 of LBOs industry but the difference disappears in CAPEX in RLBOs is consistent with
occurred 1976-87
after the IPO, as RLBOs increase
the fact that these firms being cash
CAPEX.
constrained prior to the RLBO.
839 French LBOs CAPEX increased relatively to control Increase in CAPEX is concentrated in
over 1994–2004
groups.
Private to Private deals - evidence of
existing financial constraints. In the sub
sample of Public to Private, CAPEX
reduces, but the difference to control
groups is not statistically significant.
1,009 UK buyouts Suggests that PE attempts to reorganize
Suggests that public targets suffered
from 1997 to 2006 target firms in a way which reduces
from agency whilst private targets from
inherent the targets‘ inefficiencies—
financial constraints - overinvestment
agency problems in public targets and
hypothesis in case of public firms.
investment constraints in private ones
PE exits: 345
IPO firms increase CAPEX much more
As IPO firms outperform market
SBOs and 117
than SBOs
substantially it is hard to believe that
IPOs between
they can do that while overinvesting.
2000-07
Thus the underinvestment hypothesis
seems, indirectly, more plausible.
324 Private
VC: reduction in the investment
Mentioned that firms in which leverage
Spanish firms that dependency on internal cash flows in
increased to finance the acquisition,
were subject to a SMEs in expansion stage after VC deal investments will be constrained to the
VC and PE
PE (buyouts): did not find a significant internally generated funds. Investment
Investment period sensitivity before, whereas a positive
rate falls for buyout firms.
1995–04
value is found after the acquisition.
2239 PE backed
CAPEX increases and Financial
Reduction of ICF sensitivities could be
SMEs in UK and
constraints decrease with the PE
reduction in overinvestment. However,
France spanning
intervention
effects of PE are much higher for smaller
from 1998-2007
firms - more likely to face financial
constraints and less likely to suffer
overinvestment.
1981-05 US
Investment and Capital decrease after Overinvestment interpretation is not
PE/MBO/Op.
going private.
consistent with productivity not
plant level data
changing in relation to control groups.
206 low-med tech ICF Sensitivity rises with buyout. No
PE contributes to raising target firms’
firms, o/w 108
signf. impact of buyouts on the
financing constraints and adversely
(private to private) investment rates in the post-buyout.
affect firms’ investment rates - more
buyouts 1997-04 However, results show a decrease in the likely to have lower commitments to
in FR, UK, IT and Investment of UK firms and an increase
long-term investments
SP
in the investment rates French firms.
59
Appendix II
VBA coding for Matching on Total Assets:
Sub macro_sector()
Sheets("Sector").Select
For i = 2 To 77
Sheets("Sector").Select
sector = Range("a" & i)
Sheets("PrivateTotal").Select
Range("b1").Select
Selection.AutoFilter
Selection.AutoFilter field:=2, Criteria1:=sector
Range("b2").CurrentRegion.Copy
Sheets("private").Range("a1").PasteSpecial
Application.CutCopyMode = False
Selection.AutoFilter
Sheets("PublicTotal").Select
Range("b1").Select
Selection.AutoFilter
Selection.AutoFilter field:=2, Criteria1:=sector
Range("b2").CurrentRegion.Copy
Sheets("Public").Range("a1").PasteSpecial
Application.CutCopyMode = False
Selection.AutoFilter
Call macro_match
Next
End Sub
Sub macro_match()
Sheets("Public").Select
Dim linhaspublic As Integer
linhaspublic = Range("A1048576").End(xlUp).Row
Sheets("Private").Select
Dim linhasprivate As Integer
linhasprivate = Range("A1048576").End(xlUp).Row
If linhaspublic >= linhasprivate Then
janela1 = "Private"
janela2 = "Public"
Else
janela1 = "Public"
janela2 = "Private"
End If
Sheets(janela1).Select
Do While Range("a2") <> ""
Sheets(janela1).Select
If Range("a2") <> "" Then
Rows("2:2").Select
Application.CutCopyMode = True
Selection.Cut
Sheets("Match").Select
Range("A1048576").End(xlUp).Offset(1, 0).Select
ActiveSheet.Paste
Sheets(janela1).Select
Rows(2).Delete
Sheets("Match").Select
60
linha = Range("A1048576").End(xlUp).Row
Sheets(janela2).Select
Range("Db1") = "asset ratio"
ultimalinha = Range("A1048576").End(xlUp).Row
For i = 2 To ultimalinha
If Sheets("Match").Range("f" & linha) > Sheets(janela2).Range("f" & i) Then
Range("db" & i) = Sheets("Match").Range("f" & linha) / Sheets(janela2).Range("f" & i)
Else
Range("db" & i) = Sheets(janela2).Range("f" & i) / Sheets("Match").Range("f" & linha)
End If
Next
Range("A2", "db" & ultimalinha).Select
Selection.Sort key1:=Range("db1"), Order1:=xlAscending
If Range("db2") < 2 Then
Rows("2:2").Select
Application.CutCopyMode = True
Selection.Cut
Sheets("Match").Select
Range("A1048576").End(xlUp).Offset(1, 0).Select
ActiveSheet.Paste
Sheets(janela2).Select
Rows(2).Delete
Else
Sheets("match").Select
apagar = Range("A1048576").End(xlUp).Row
Rows(apagar).Delete
End If
End If
Loop
Sheets("Private").Select
Cells.Select
Selection.ClearContents
Sheets("Public").Select
Cells.Select
Selection.ClearContents
End Sub
For matching on ROA the approach was the same, with exception of the calliper based
criteria, in which we substitute the ratio between the max. (TA) / min. (TA) < 2 (the part
in bold) with forcing the difference between ROA’s to be less than 2x the min. (ROA):
VBA coding adjustment (replace part in bold) for Matching on ROA:
If Sheets("Match").Range("f" & linha) > Sheets(janela2).Range("f" & i) Then
Range("db" & i) = Abs(Sheets("Match").Range("f" & linha) - Sheets(janela2).Range("f" & i))
Else
Range("db" & i) = Abs(Sheets(janela2).Range("f" & i) - Sheets("Match").Range("f" & linha))
End If
Next
Range("A2", "db" & ultimalinha).Select
Selection.Sort key1:=Range("db1"), Order1:=xlAscending
If Range("db2") < 2 * WorksheetFunction.Min(Sheets(janela2).Range("f" & i),
Sheets("Match").Range("f" & linha)) Then
61
Appendix III
Table 13 - Public and Private firms per Country/Legal Form
This Table reports the country and legal form breakdown of our sample of
public and private companies, before and after each matching procedure.
Country ISO / Legal Form
AT
Private
BE
Private
Public
BG
Private
Public
CZ
Private
Public
DE
Private
Public
ES
Private
Public
FI
Private
Public
FR
Private
Public
GB
Private
Public
GR
Private
Public
HU
Private
Matched on NACE 4-digit plus:
TA
ROA
TA & ROA
All
22
20
19
22
20
19
19
836
626
657
574
67
769
59
567
41
616
49
525
158
97
108
83
101
57
46
51
60
48
39
44
449
310
313
277
302
147
188
122
192
121
170
107
461
375
307
336
355
106
294
81
219
88
264
72
1,411
1,015
1,099
947
518
893
451
564
377
722
423
524
415
258
265
240
387
28
240
18
242
23
223
17
8,375
5,861
6,005
5,195
2,906
5,469
1,260
4,601
1,520
4,485
1,118
4,077
1,172
946
760
826
1,029
143
856
90
652
108
748
78
854
730
726
667
20
834
11
719
11
715
10
657
68
49
44
42
67
49
44
42
12,444
6,591
7,186
6,001
9,566
2,878
4,326
2,265
4,751
2,435
3,911
2,090
6
5
4
5
4
2
4
1
3
1
4
1
183
144
137
132
144
39
114
30
106
31
105
27
173
124
127
113
25
148
23
101
20
107
22
91
2,176
1,311
1,266
1,192
2,133
43
1,277
34
1,235
31
1,165
27
233
150
169
137
158
75
88
62
104
65
81
56
29,437
18,612
19,192
16,786
IE
1
Public
IT
Private
Public
1
LV
Private
Public
PL
Private
Public
PT
Private
Public
SE
Private
Public
SK
Private
Public
Total
19
62